# ERPClaw: Full Content Export

> The AI-native ERP for the QuickBooks-NetSuite gap. Run the whole business in one system, with 14 industry verticals. Free and open source, self-hosted.
> Machine-readable full text of 160 published pages, generated at build time from https://www.erpclaw.ai. Canonical source is the live site; each entry links its URL.

---

## AI Accounting Software, Free Forever, $0
URL: https://www.erpclaw.ai/ai-accounting/

# AI Accounting Software, Free Forever
The only AI accounting that's AI-native AND open-source AND self-hosted AND $0. Run it on your laptop. Read the source.
covers every major business function·7 currencies·fully open source
Try the guided chat demo Install in 5 minutes
Live on the Stripe Marketplace. Real production users.

## What is AI accounting?
AI accounting is accounting software where an AI agent reads, writes, and reconciles your books directly. In an AI-native tool the agent posts journal entries, runs reports, and closes periods on its own. In AI-decorated tools the AI only suggests; a human still drives every meaningful write. ERPClaw is the only AI-native open-source AI accounting tool, and it is free forever.

- ·AI-native, not AI-decorated: the agent posts double-entry journals end to end; it does not just summarize the dashboard.
- ·Open source: fully Free & Open Source. Audit the code, fork it, run it.
- ·Self-hosted: SQLite by default; PostgreSQL fully supported. Your books live on your machine, not a SaaS vendor.
- ·Free forever: $0 software cost. No subscription, no per-seat pricing, no metered usage.
- ·Production-ready: extensive automated test coverage, immutable general ledger, ASC 606 revenue recognition, 7 currencies.

## Ask it like you'd ask a person
Real requests, each backed by shipped, tested capability. You ask; real accounting code does the work.
"Chase the late payers for me"
Overdue invoices get reminder emails automatically, on the schedule and wording you set. The emails really go out.

"The customer paid $50 short; write it off in the same payment"
One request records the payment and the write-off together, posted and balanced, no second trip through the books.

"Which invoices are due this week?"
The list, from your books: who, how much, and how overdue, ready to act on.

"How much sales tax did we collect last quarter?"
US sales tax collected and owed over any date range, straight from the books. ERPClaw reports your tax; it does not file it, and your accountant stays your filer.

## Most AI accounting tools don't actually do accounting
Open any "AI accounting" product page in 2026 and you'll see a chat box on top of a SaaS subscription. You ask a question. The AI suggests an answer. You click "approve" and a human-designed workflow runs underneath. The AI never touches your books. It cannot post a journal entry on its own. It cannot reconcile a Stripe payout end to end. It cannot run a trial balance unless somebody built that exact button. It just summarizes what you already see in the dashboard.
That works if you already know accounting and want a faster search box. It does not work if you're a founder running 30 invoices, 4 bank accounts, and a Stripe sync at midnight, hoping the books are right by tax season. The deeper bookkeeping cut is at AI bookkeeping; if your QuickBooks data has been stuck in this exact loop, the switching guide is at AI for QuickBooks.

- AI suggestions still need a human to click "post"
- Your books live on a vendor's server, not your machine
- Subscription pricing scales with seats you don't have
"Ask any AI accounting tool one question: can it post a journal entry to your general ledger without a human clicking approve? Most cannot. They classify, they suggest, and they hand the work back. That is autocomplete on a bookkeeping form, not accounting."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## What "AI-native" actually means
Every accounting vendor in 2026 says their product is AI-powered. The architecture tells you whether that means anything. There are two camps.
AI-decorated

### The AI is a sidecar
QuickBooks, Xero, Sage, NetSuite, Vic.ai. The schema, the GL rules, and the write paths were all designed before AI existed. The AI was added later as a chat assistant. It reads. It suggests. A human approves. The AI cannot post a journal entry without human consent because the platform was never designed for that.
Tell: the AI lives in a sidebar. Every action ends in "review and approve."

AI-native

### The AI is the primary user
ERPClaw. The data model, the 12-step GL validation, and the action layer were designed for an AI agent to use. Every action is a single transaction with full rollback. The AI doesn't suggest. It submits. The invariants run before any posting touches the books, so the AI cannot violate the GL even when it's wrong about something else.
Tell: chat is the only interface. Every action is a verb the AI can call directly.

Once the AI does the operating and the books check themselves, the rest of this page follows naturally. Books update in real time. The month closes itself. You ask a question and get a number, not a suggestion.
For the broader architecture argument across the full ERP, see the AI-native ERP pillar. For the open-source angle and how ERPClaw compares against ERPNext add-ons and Odoo's automation layer, see open-source AI accounting. For the commercial-intent buyer's guide with vendor matrix, see AI ERP 2026.

## What AI accounting does in ERPClaw

### Submit invoices in plain English
Type "Create an invoice for Wayne Enterprises, $12,500 for consulting services, net 30" and ERPClaw drafts the invoice, picks the right GL accounts, and waits for your submit. Submit it and the GL entries post in one transaction.

### Reconcile Stripe payouts automatically
Stripe pays you one lump sum with fees, refunds, and disputes already taken out. ERPClaw splits each payout back into its real pieces, into clearing-account entries so your cash, fees, and revenue land in the right accounts. End the month with $0 in the clearing account, every time.

### Run financial reports on demand
"Show me Q1 P&L by segment." "Trial balance as of March 31." "Customer aging older than 60 days." The AI runs the query against your real ledger and returns a clean answer in seconds. No SQL, no menu hunting.

### Catch GL errors before they post
Every journal entry runs through invariant checks before it touches the books. Debits equal credits, accounts exist, the period is open, the company is right, no future-dated postings. If a check fails, the transaction rolls back. Bad entries cannot reach your GL.

### Multi-currency invoicing across 7 currencies
Bill in USD, EUR, GBP, CAD, INR, SGD, or AED. Invoice currency must equal payment currency, no FX conversion in our books. Stripe handles the conversion if your customer pays from a different country. Your books stay clean.

### Audit trail on every AI action
Every action the AI takes is logged with full input, full output, and the exact GL entries that posted. You can replay any decision, audit any close, and explain any number to your accountant. The AI cannot do anything you cannot reconstruct later.

## ERPClaw vs the AI accounting field
We picked the five most-cited AI accounting products in 2026 SERPs. Three are AI-native but proprietary and paid. Two are AI-decorated. ERPClaw is the only one in all five boxes that matter.
| Feature | ERPClaw | Puzzle | QuickBooks + Intuit Assist | Xero + Just Ask Xero | Sage + Copilot | Vic.ai
| AI-native architecture | Yes, from commit one | Free tier | No (chat layer) | No (chat layer) | No (taken offline 2025) | Partial (AP only)
| AI writes to the GL directly | Yes, every action | Partial | Suggestions only | Suggestions only | Suggestions only | Partial
| Open source | open source license | No | No | No | No | No
| Self-hosted | Your machine | Cloud only | Cloud only | Cloud only | Cloud only | Cloud only
| Pricing | $0 forever | Free tier + paid | $30 to $200 / mo | $15 to $78 / mo | $11 to $79 / user / mo | Enterprise
| Multi-currency | 7 currencies | Limited | Plus tier and up | Established and up | Yes | Limited
| Real-time GL validation | Full invariant check | Manual | Manual | Manual | Manual | Not applicable
| Vendor lock-in | None | High | High | High | High | High

Cloud AI accounting assistants have run into data-segregation incidents in shared-tenancy setups. Self-hosted accounting eliminates that class of risk. Your books are on your machine. For mid-market head-to-heads see ERPClaw vs NetSuite, vs Sage Intacct, and vs Rillet. If you're weighing the open-source and SMB field instead, compare ERPClaw as an Odoo alternative, an ERPNext alternative, and against Xero for AI accounting.

## Who this is for

### Solo founder
You're invoicing customers, paying contractors, and reconciling Stripe at midnight. You don't want a $50/month subscription that scales with seats you don't have.

### Shopify store owner
Stripe and Shopify both pay you net of fees. You want one system that posts the right entries, tracks COGS, and tells you what you actually made this month.

### Small CPA practice
Your clients want AI in their books. You want auditability, immutable GL, and a system you can read the source of. ERPClaw is the only one that hits all three.

### SaaS startup
ASC 606 deferred revenue, MRR, and Stripe billing all need to flow into your books cleanly. ERPClaw's Stripe addon (live on the Stripe Marketplace) handles all three.

## Pricing
$0 forever
open source. Self-hosted. Unlimited users. No per-transaction fee. No upgrade tier.
QuickBooks Plus is $85 per month. Over five years, that's $5,100. ERPClaw is $0, runs on your laptop, and the AI doesn't need a separate add-on subscription. The deeper architecture argument for why open source matters here is at open-source AI accounting, and the foundation that makes AI-native possible is at features/foundation.
See full pricing details

## Why this is real
GL invariants on every posting

industry verticals shipped

currencies supported

ASC 606
Stripe revenue recognition

Free
and open source, forever

Stripe
Marketplace listed and selling

Patent pending·Trademark filed·Stripe Marketplace listed·github.com/avansaber/erpclaw
"We put the whole ledger on your own machine under GPL v3 because the alternative is asking a business to trust its books to a shared tenancy it cannot inspect. When the software is free and the data never leaves your server, the only thing left to argue about is whether the accounting is correct."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## Common questions about AI accounting

### What is AI accounting software?
AI accounting software uses machine learning and language models to handle the work that used to take a bookkeeper or accountant hours. The strong versions categorize transactions, reconcile bank and Stripe payouts, draft invoices from a sentence, and run financial reports on demand. The weak versions just summarize what you can already see in your dashboard.

### How does AI accounting software work?
You connect your accounts (bank, Stripe, Shopify), the AI reads the transactions, and it either suggests categorizations for a human to approve or posts them directly with confidence checks. ERPClaw is in the second camp: the AI is the primary user of the system, every action runs through the full set of GL invariants, and the books update in real time.

### Is AI accounting software accurate?
It depends on the architecture. Vendors that bolted AI on top of a legacy database typically claim 90-97% categorization accuracy. ERPClaw runs every posting through the full set of GL invariants (debits equal credits, accounts exist, period is open, no future dates) before it touches the books, so a posting either passes all checks or rolls back. Confidence is checked, but invariants are enforced.

### Is there free AI accounting software?
Wave is free but does not have AI architecture. Puzzle has a free tier but is closed-source and cloud-only. ERPClaw is free forever, open source, runs on your laptop or your own server, and is AI-native end to end. There is no per-user fee, no per-transaction fee, and no upgrade tier.

### Does AI accounting software integrate with Stripe?
Yes, ERPClaw has a deep Stripe integration that's live on the Stripe App Marketplace. It auto-syncs every charge, refund, and payout, posts GL entries for each transaction, reconciles payouts to the bank in one command, and tracks ASC 606 deferred revenue. Most other AI accounting tools rely on third-party connectors like Synder or A2X; ERPClaw is native.

### What is the difference between AI accounting and automated accounting?
Automated accounting uses fixed rules. "If the merchant is Stripe, post to category Stripe Fees." AI accounting learns from context. The AI reads the memo, the amount, the vendor history, and decides what category fits. Automated misses anything new. AI handles the new.

### Can AI replace my accountant?
Not the judgment work. AI handles the daily posting, reconciliation, and report generation that used to fill an accountant's afternoon. Your accountant still owns the judgment calls, tax strategy, and audit defense. The hybrid model (AI does the volume, human does the strategy) is where the math actually works.

### How does ERPClaw compare to QuickBooks for AI accounting?
QuickBooks Online by Intuit added Intuit Assist on top of a database designed in the early 1990s. The AI suggests; a human still drives every meaningful posting. ERPClaw is AI-native: the agent posts double-entry journals end to end with invariants enforced at write time. Pricing is the other half. QuickBooks Plus is $85/month (about $5,100 over five years); ERPClaw is $0 forever. The full head-to-head is at ERPClaw vs QuickBooks and the QuickBooks-specific switching guide is at AI for QuickBooks.

### Is ERPClaw really comparable to NetSuite or Sage Intacct for mid-market accounting?
On the primitives, yes: immutable double-entry GL, multi-entity, ASC 606, multi-currency, audit trail. ERPClaw also covers inventory, manufacturing, payroll, Stripe, and Shopify natively. NetSuite still leads on multi-entity consolidation polish at scale; Sage Intacct still leads on ASC 842 lease accounting and AICPA preferred-provider status. The full breakdowns: ERPClaw vs NetSuite and ERPClaw vs Sage Intacct.

### Does AI accounting work with my inventory and bookkeeping in the same system?
Yes, and that's the point of one shared database. The same AI that posts your Stripe charges also reconciles your bookkeeping, runs inventory valuation, and posts stock movements to the GL. There's no second system to sync. The bookkeeping deep-dive is at AI bookkeeping and the inventory deep-dive at AI inventory.

### What does "AI-native" actually require in the foundation?
Five things. The AI runs the system itself, not a chat box glued onto menus. The data was organized for the AI to work in from day one. Stripe and Shopify connect directly, with no middleman tool. The books are checked the moment anything is written, not after the fact. And the code is open, so anyone can verify all of it. ERPClaw's foundation hits all five. More on the architecture at features/foundation and the deeper open-source argument at open-source AI accounting.

### How long does it actually take to install and start using ERPClaw?
5 minutes for the install (4 commands: install, init, connect Stripe, run-server). 15 to 30 minutes more to import your chart of accounts and opening balances if you're migrating from QuickBooks, Xero, or another system. By the end of an hour, the AI agent is posting journal entries against your live data. Install steps are at docs/core/install.

---

## AI Bookkeeping Software: Free, Open Source, Self-Hosted
URL: https://www.erpclaw.ai/ai-bookkeeping/

# AI bookkeeping that actually books entries
Most "AI bookkeeping" tools are a chat box on top of a $400-per-month bookkeeper subscription. ERPClaw is AI-native end-to-end. The agent categorizes, reconciles, accrues, and reports without a human re-keying anything. Open source, self-hosted, $0 forever.
Try the demoInstall in 5 minutes

## What is AI bookkeeping?
AI bookkeeping is bookkeeping software where an AI agent does the daily transaction work end-to-end: categorize, post, reconcile, accrue, report. AI-native bookkeeping has no human in the loop for the volume work; hybrid AI bookkeeping (Botkeeper, Pilot, Docyt, Bookkeeper360) keeps a human reviewer because the AI is bolted onto a workflow that still expects one. ERPClaw is the AI-native open-source option.

- ·No human bookkeeper subscription. The AI does the volume; your CPA still owns judgment calls.
- ·Open source. Audited Python under Free & Open Source. Read the code, fork it, fix it.
- ·Self-hosted. Your books live on your hardware. SQLite default; PostgreSQL fully supported.
- ·Immutable GL. Cancel = reverse, never edit. Audit-trail integrity is structural, not a setting.
- ·$0 forever. No SaaS tier, no per-seat fee, no metered usage.

## Ask it like you'd ask a person
Bookkeeping requests that work today, in plain words, on statement files your bank already exports.
"Import my January bank statement and match it to open invoices"
Hand it the statement file (OFX, CAMT.053, MT940, or BAI2). Lines that match themselves are done; the rest queue for your review.

"What is still unmatched on that statement?"
The short list a human should actually look at, with one-line fixes: match it to Acme, map it to an account, or leave it.

"How does the checking account reconcile as of month end?"
The as-of-date picture from your imported statements: matched, unmatched, and where the account stands.

"The customer paid $50 short; write it off in the same payment"
Payment and write-off posted together, balanced, in one request.

## What AI bookkeeping does in ERPClaw

### Transaction categorization
The AI agent reads your bank statement file, Stripe payouts, and Shopify orders, then categorizes each transaction against your chart of accounts. Categorization rules learn from your prior decisions; new patterns get flagged for review, not silently posted.

### Recurring entries on autopilot
Subscriptions, rent, payroll, depreciation, prepaid expense amortization. Each runs on its own schedule; the AI posts the journal entry on the right day with the right offset accounts, all inside ERPClaw's immutable GL.

### Bank and credit card reconciliation
The agent matches every imported bank transaction against the corresponding ledger entry. Unmatched items surface as a queue you actually clear instead of a pile you ignore. Stripe payouts reconcile against invoices automatically.

### Month-end accruals
Standard accruals (deferred revenue, prepaid expense, accrued liability) post per the close calendar. The agent computes the value, posts the entry, and reverses next period when the rule says to.

### ASC 606 revenue recognition
The Stripe integration handles performance-obligation tracking, transaction-price allocation, and recognition over time or at point in time. ASC 606 stops being a quarterly engagement and becomes a daily background process.

### Reports and trial balance
P&L, balance sheet, cash flow, AR aging, AP aging, GL detail, trial balance, sales tax summary, 1099 summary. The core financial report set out of the box. The agent generates the report; you review or export.

### Multi-entity bookkeeping
Run several companies in one shared database, each with its own immutable GL. Intercompany transactions mirror across entities, and reporting rolls up per entity or across the group. No separate instance, no per-entity license fee.

## AI bookkeeping vendor comparison
| Vendor | AI posture | License | Self-host | Cost
| Botkeeper | Hybrid (AI plus human bookkeepers) | Proprietary | No (vendor cloud) | Founder-reported $200-$1,500+/month per client
| Pilot | Human bookkeeping with AI assist | Proprietary | No | $499+/month per company
| Docyt | Workflow automation + AI categorization | Proprietary | No | $299-$1,499/month
| Bookkeeper360 | Hybrid SaaS plus advisory | Proprietary | No | $129-$549+/month
| ERPClaw | AI-native end-to-end (no human required) | Free & Open Source | Yes (SQLite or PostgreSQL) | $0 forever

## Who runs AI bookkeeping on ERPClaw
Three profiles where the AI-native open-source flow beats hybrid SaaS today. Each one is hosting a real combination of Stripe payouts, Shopify orders, and recurring entries that hybrid services charge $200 to $1,500 a month to handle.

### Stripe-only SaaS founders
Stripe is your only payment processor. Subscriptions, one-off charges, refunds, disputes, and ASC 606 deferred revenue all need to land in the books cleanly. The Stripe integration (live on the Stripe Marketplace) backfills history and auto-reconciles payouts. The Stripe docs are at /docs/stripe/.

### Shopify store operators
Shopify orders, payouts net of fees, returns, and COGS all need to flow into the GL with the right cost basis. The Shopify integration (OAuth Token Exchange, v1.1.3) pulls orders, payouts, and adjustments directly. Inventory and bookkeeping live in the same database, so COGS is correct on every shipment. Connected inventory deep-dive at /ai-inventory/.

### QuickBooks switchers
Your books are stuck in QuickBooks with manual Stripe reconciliation eating hours every month. The AI-native upgrade is at AI for QuickBooks with the full head-to-head at ERPClaw vs QuickBooks and the weekend migration playbook at /migrate/from-quickbooks/.

For the broader category framing see AI accounting and the architecture argument at open-source AI accounting. For mid-market consolidation comparisons see vs NetSuite, vs Sage Intacct, and vs Rillet.

## Frequently asked questions about AI bookkeeping

### What is AI bookkeeping?
AI bookkeeping is bookkeeping software where an AI agent categorizes transactions, posts journal entries, runs reconciliations, and produces reports without a human re-keying anything. AI-native bookkeeping (like ERPClaw) does this end-to-end with the agent owning the workflow. Hybrid AI bookkeeping (like Botkeeper, Pilot) keeps a human bookkeeper in the loop for judgment and review.

### Can AI really do my bookkeeping without a human?
Yes for the volume work. AI agents are reliable at categorizing transactions, posting recurring entries, reconciling bank statements, and generating reports. They are still imperfect at judgment calls (unusual transactions, ambiguous categorization, audit defense). The right model is AI does the volume, human does the strategy. ERPClaw flags ambiguous items for review instead of silently posting.

### How much does AI bookkeeping cost?
It varies by model. Hybrid services (Botkeeper, Pilot, Bookkeeper360, Docyt) charge $200 to $1,500 per month per client because a human reviews everything. Pure-AI tools are cheaper but rare. ERPClaw is $0 forever (open source, self-hosted) with the AI doing the work; you only pay for hosting infrastructure.

### What about my CPA, do I still need them?
Yes. AI bookkeeping handles daily volume, not tax strategy or audit defense. Your CPA still owns year-end review, tax planning, and entity-level decisions. The hybrid model is AI does the volume your CPA was billing you for at $200 per hour, your CPA does the judgment work that actually requires their license. Most teams come out ahead on both cost and quality.

### Is open-source AI bookkeeping safe for my finance data?
Self-hosted ERPClaw means your books live on your hardware in a SQLite file or PostgreSQL database you control. Nothing leaves your infrastructure unless you explicitly send it. The AI agent runs locally or in your cloud. By contrast, every SaaS bookkeeping service stores your data on vendor servers; you trade local control for vendor convenience.

### What if the AI miscategorizes a transaction?
ERPClaw's immutable GL means you cannot silently overwrite a posting. To fix a miscategorization, the agent posts a reversal entry plus the correct entry, both with full audit trail. Categorization rules learn from corrections; the next similar transaction goes to the right account on its own.

### Will my CPA accept AI-generated books for tax filing?
Yes. ERPClaw exports trial balance, P&L, balance sheet, and GL detail in standard formats (CSV, PDF). The chain-hashed audit trail and immutable GL design produce books that pass external review cleanly. CPAs increasingly prefer this style of structured, auditable books over mutable-history alternatives.

Related: read the broader AI accounting story, the AI-native ERP framework, the QuickBooks comparison, the QuickBooks migration playbook, the nonprofit view if you keep restricted funds, or the double-entry bookkeeping for developers deep-dive.

---

## AI CRM That Follows the Deal Into the Books
URL: https://www.erpclaw.ai/ai-crm/

# Most CRMs stop at closed-won. That is where the work starts.
You win the deal, and then somebody retypes it into the system that does invoicing. Then somebody checks whether the stock exists. Then somebody chases the payment in a third place. ERPClaw keeps going: the won opportunity becomes a quote, an order, an invoice, and cash, with the books moving underneath the whole way. Free, self-hosted, open source.

## One deal, all the way to the money
This is the whole argument for the page, so here it is in order. Every step happens in the same system, against the same records.

### A deal is won
The opportunity closes in the pipeline your team already works in.

### It becomes a quotation
The won opportunity converts straight into a quotation. Nobody retypes the customer, the items, or the numbers into a second system.

### Then an order and a delivery
The quotation becomes a sales order, stock is reserved against it, and the delivery draws from real inventory rather than a promise.

### Then an invoice
The invoice is raised from the order, so what you billed and what you shipped are the same document lineage.

### Then cash
The payment lands against that invoice, with short payments written off in the same step if the customer settles light.

### And the books move
Every one of those steps posts real double-entry accounting as it happens, checked after each posting. The deal and the ledger are the same story, not two systems reconciled monthly.

Nothing in that sequence involves an integration, a sync, or a nightly job between two vendors. It is one database, and the accounting ischecked after every posting.

## What the sales side does

### The pipeline, shaped like your business
Leads, contacts, companies and opportunities, moving through stages you define rather than a fixed list someone else picked. An agency going brief to proposal to retainer and a manufacturer going enquiry to sample to tooling to production both get their own pipeline, and the reports follow the stages you actually use.

### Every follow-up attached to something real
Tasks and activities hang off the lead, the deal, the contact or the company they belong to, with owners and due dates. A task cannot point at a deal that does not exist, which is how orphaned reminders quietly accumulate in looser systems.

### The lists you actually reuse, saved
Deals stuck in proposal. Leads from the trade show nobody has called. Save the filter once, replay it by name, and search across every sales record when you cannot remember which kind of thing you are looking for.

### Territories, quotas and performance
Assign territories, set quotas, and read pipeline, funnel, velocity and win-loss reporting off the same records your salespeople work in, not a spreadsheet someone rebuilds every Friday.

### Campaigns that end at the ledger
Campaigns and recipient lists live in the same system as the invoices they eventually produce, so the spend and the revenue are not in two different tools that disagree.

### Bring your existing CRM across
Leads, contacts, companies and deals import from CSV, the format every CRM exports, with duplicate handling you choose up front rather than a silent default. No wizard fee, no export ransom.

## About the other AI CRMs
They are real, and they are good at what they do. Attio runs a fleet of named agents across scoring, routing, enrichment, briefing and churn risk, metered by credits. Twenty is open source and self-hostable with a chatbot, workflow agents and a native MCP server. If you want the best front office on the market, look at them properly rather than taking our word for anything.
What neither can do is follow the deal past the handshake, because there is no ledger underneath them to follow it into. That is not a criticism of their engineering. It is a description of where a front-office product ends. If the expensive part of your week happens after the deal is won, that is the part we built.

## Switched on per business
These are set up for your business rather than toggled on in a settings page, because each one touches something specific to you: your mailboxes, your scoring rules, your attribution model. Ask below and we set yours up.

### Email and calendar capture
Customer conversations attach themselves to the right account and opportunity, with source, timestamp and owner recorded, and opt-out controls that respect who is on the call.

### Call and meeting notes
A recording or transcript becomes a reviewable summary with decisions and next steps, linked to the deal, without touching a financial record on its own.

### Lead scoring that runs itself
Scores update from what actually happened rather than a manual review, and show you the evidence behind the number.

### Nurture sequences that fire
Sequences run on their own schedule against the lists you built, instead of sitting as definitions waiting for someone to press go.

### Next best action and deal risk
A recommendation that names its evidence, its owner and its due date, and turns into a real task the moment you accept it.

### Revenue attributed back to the deal
Trace a paid invoice back through the opportunity and the campaign that started it, with the attribution model stated rather than assumed.

## Common questions

### What makes a CRM AI-native rather than a CRM with AI in it?
Most CRMs added AI to the front office: better summaries, better drafting, better routing. That is genuinely useful and the good ones do it well. AI-native means the assistant is the primary way the system is operated, and it can act rather than only suggest, because every action in the system is callable in plain language. The test is not how good the chat is. It is whether the thing it decides actually posts.

### How is this different from Attio or Twenty?
Both ship real AI and both are worth looking at. Attio runs a set of named agents across scoring, routing, enrichment, briefing and churn risk, and sells them credit-metered. Twenty is open source, self-hostable, and ships a chatbot, workflow agents and a native MCP server. The difference is not who has AI. It is where the software stops. Both are front-office systems: the deal closes and the work moves to whatever runs your invoicing, inventory and books. In ERPClaw the deal keeps going through quote, order, delivery, invoice, payment and the ledger, because it is one system.

### Do I have to run the whole ERP to use the CRM?
No. Install it and use the sales side on its own if that is all you need today. The rest of the system is already underneath, which is the point: the day you want the deal to become an invoice, nothing has to be integrated, migrated or bought.

### Can it replace HubSpot or Salesforce for a small team?
For a lot of small teams, yes, and honestly for some it will not. If your business runs on marketing automation depth, a large app marketplace, or a partner ecosystem, those products have a decade of surface you would miss. If your pain is that the deal dies at closed-won and someone re-keys it into accounting, that is exactly the gap this closes.

### Where does my data live?
On your hardware. ERPClaw is self-hosted and open source, so customer records, deal history and the books behind them stay on a machine you control, in a database you can read.

### What does it cost?
Nothing. ERPClaw is free and open source, self-hosted, and the sales side is not a paid tier, a seat count, or an add-on.

Related: the full ERP the pipeline sits inside,Insights if you want answers before you want a system of record, and the wider feature set.

## We set this up for you, one business at a time.
Tell us what your sales process looks like and which of the pieces above you want connected to your mailboxes and your books. We set these up one business at a time.

---

## AI ERP 2026: open source, AI-native, free
URL: https://www.erpclaw.ai/ai-erp/

# AI ERP: open source, AI-native, free
What AI ERP actually is in 2026. How to evaluate AI-native architecture vs AI-decorated bolt-ons. Why ERPClaw is the only open-source AI-native option, and the honest gaps in every other AI ERP claim.
Every module·One shared database·14 verticals·$0 forever
Try the demo See the 13-vendor matrix

## What is AI ERP?
AI ERP is enterprise resource planning software where an AI agent reads, writes, and reconciles your data directly. The agent posts journal entries, runs reports, and closes periods on its own with human governance. AI ERP is the commercial category; AI-native ERP is the architectural sub-category that actually rebuilds around the agent. Most products marketed as AI ERP in 2026 are AI-decorated bolt-ons, not AI-native.

- ·AI-native: ERPClaw, Rillet (finance only). Architecture rebuilt around the agent.
- ·AI-decorated: NetSuite Joule, Sage Copilot, QuickBooks Intuit Assist, Xero Just Ask, Odoo plug-in stack, ERPNext add-ons. Chat sidebar on a forms-and-workflows product.
- ·Open source AND AI-native: ERPClaw only.
- ·$0 forever: ERPClaw is the only AI ERP at this price point. Everything else is seat-based or contact-sales.

## Why AI ERP matters in 2026
Every ERP since SAP R/3 in 1992 follows the same recipe: forms capture data, workflows route approvals, batch jobs reconcile and post, BI dashboards summarize the result. The recipe is mature and battle-tested. It also assumes a human is driving every meaningful write to the database.
That assumption breaks in 2026. AI agents can read, write, reconcile, and reason. The bottleneck is no longer the human clicking through forms. The bottleneck is the architecture's willingness to let the agent be the primary user. AI-native ERP makes that change at the architecture layer; AI-decorated ERP papers over it with a chat sidebar.
The structural consequence: AI-native vendors can ship features 10x faster because the AI is part of the build pipeline. AI-decorated vendors are still maintaining 20-year-old codebases plus a copilot. Over 3 to 5 years, the gap compounds.

## Who AI ERP is for

### Mid-market CFO
Pain: Two FTEs on NetSuite admin. 30 to 40 percent renewal increase. Quarterly close is a fire drill.
Fit: Full ERP scope, $0 license, AI-native architecture. Cancel NetSuite at renewal; migration is 2 to 6 weeks.

### Founder running Stripe + Shopify
Pain: QuickBooks Online sync drops orders. Manual fee reconciliation eats 10 hours a month. Hit the user cap.
Fit: First-party Stripe (every operation, ASC 606) + Shopify (every operation). One ERP, one database, no manual fee work.

### Open-source CTO
Pain: ERPNext install is a 6-hour dance. Odoo Community gates the apps you actually want. AI is plug-in stack.
Fit: Free & Open Source, 5-minute install, AI-native by design. Spec-first regen. No partner dependency.

## What AI ERP actually does

### Conversational chart of accounts setup
An AI agent sets up your full US GAAP chart of accounts in one prompt. ERPClaw's setup-chart-of-accounts action seeds the standard template; you adapt by chatting, not by clicking through 12 screens.

### Daily auto-reconciliation across channels
Stripe payouts, Shopify orders, and bank deposits reconcile automatically every day. The 12-step GL validation runs on every transaction. Manual fee reconciliation drops from 10 hours a month to zero.

### AI-driven month-end close
The agent runs accruals, posts deferred-revenue rollforwards, generates trial balance, and surfaces variance against budget. The CFO reviews exceptions, not entries. Close timeline shrinks from a fire drill to a Monday morning.

### Self-improving module generation
ERPClaw OS includes a sandboxed module-generation engine. Your finance team describes a missing feature in plain English; the AI generates a new SKILL.md spec, runs sandbox tests, and proposes the module for review. Customization is no longer an engagement.

## AI ERP vendor comparison
| Vendor | Architecture | License | Self-host | Cost
| NetSuite | AI-decorated (Joule on 1998 ERP) | Proprietary | No | $30K to $500K all-in
| Sage Intacct | AI-decorated (Sage Copilot) | Proprietary | No | $15K+/year + 6-fig implementation
| Odoo | AI-decorated (plug-in stack) | Open-core (LGPL Community + proprietary tiers) | Yes (Community); No (Online) | Community free; Standard $31.10/user/mo
| ERPNext | AI-decorated (changAI, NextAI, Ollama, Composio) | Free & Open Source | Yes | Free self-host or Frappe Cloud SaaS
| ERPClaw | AI-native by design | Free & Open Source | Yes (SQLite or PostgreSQL) | $0 forever

Full vendor matrix, with per-vendor head-to-head pages linked from it, at /erp-comparison/.

## Common objections, honest answers

### Doesn't every ERP have AI now?
Yes, every ERP has a Copilot or chat sidebar branded as AI. None of them rebuilt the architecture. The 5-trait test on our AI-native ERP framework separates AI-native architectures from AI-decorated bolt-ons.

### How can ERPClaw be free if it's a real ERP?
ERPClaw is open-source under Free & Open Source. The software is $0 forever because there is no SaaS company between you and the code. AvanSaber Inc. (the company behind ERPClaw) makes money from optional implementation services and enterprise support contracts; the product itself is free.

### What about audit and compliance for open-source ERP?
ERPClaw enforces constitutional articles at write time, including immutable GL (cancel = reverse, never edit), 12-step posting validation, double-entry balance, and chain-hashed audit trail. SOC 2-style controls are documented. Auditors increasingly prefer observable open-source controls over SaaS black boxes.

### Is AI ERP ready for a real business or still a toy?
ERPClaw is on the Stripe Marketplace with paying customers. Layered automated tests, constitutional tests, invariant checks. ASC 606 revenue recognition, immutable GL, US tax forms (W-2, 1099, NACHA, FICA, FUTA, SUTA). Real production. Not a toy.

### What's the catch?
ERPClaw's web dashboard is included free and currently in beta: it builds its screens for every installed module automatically, and it runs alongside chat, which stays the fastest way to work with Stripe and Shopify today. The primary interface is the AI agent. Support is community-tier (GitHub, Discord) plus optional paid services. Multi-thousand-user Fortune 500 deployments are not our target. Within those constraints, the architectural moats compound year over year.

## Frequently asked questions about AI ERP

### What is AI ERP and how is it different from regular ERP?
AI ERP is enterprise resource planning software where an AI agent reads, writes, and reasons over the same data layer that humans do. Regular ERP routes a human through forms and approvals; AI ERP lets an agent do the same routing autonomously, with governance gates. ERPClaw is AI-native by architecture: every business action (post-gl-entries, add-customer, reconcile-stripe-payout, and 3,124 more) is invokable by an agent without UI clicks.

### What's the best AI ERP in 2026?
It depends on your scope. ERPClaw is the only open-source AI-native option with full ERP scope and $0 forever pricing. Rillet is the strongest commercial AI-native finance close SaaS but covers finance only. NetSuite, Sage Intacct, Odoo, and ERPNext are AI-decorated, not AI-native. See our 5-trait test to make the decision yourself.

### Can AI ERP replace my accountant?
Not the judgment work. AI handles the daily posting, reconciliation, and report generation that used to fill an accountant's afternoon. Your accountant still owns judgment calls, tax strategy, and audit defense. The hybrid model (AI does the volume, human does the strategy) is where the math actually works.

### How much does AI ERP cost?
It varies wildly. NetSuite + Joule starts around $30K per year and scales with users and modules. Rillet is seat-based at founder-reported $2K to $10K per month. Sage Intacct + Copilot is $15K+/year with add-on modules priced above the core. ERPClaw is $0 forever; the only ongoing cost is hosting infrastructure (typically $0 to $20 per month).

### Is AI ERP secure for finance data?
ERPClaw is self-hosted on your infrastructure, so the security perimeter is yours. The AI agent runs locally or in your cloud; no data leaves your environment unless you explicitly send it. SaaS AI ERP (NetSuite, Rillet, Sage Intacct) sends your data to vendor cloud where the security perimeter is theirs. Both models can be SOC 2 compliant; pick based on whose risk model you trust.

### Can I migrate from QuickBooks or NetSuite to an AI ERP?
Yes. ERPClaw has migration guides for QuickBooks (a weekend playbook) and NetSuite (a mid-market playbook with SuiteScript translation). The AI agent does the data transcription; you spot-check trial balance and aging, then run parallel for one to two weeks before cutover. See /migrate/ for guides.

### What's the difference between AI ERP and AI-native ERP?
Buyers ask both questions. AI ERP is the commercial term covering any ERP with AI features. AI-native ERP is the architectural term for ERP designed around an AI agent from line one, not retrofit. The 5-trait test on our AI-native ERP page separates the two. Most products marketed as AI ERP today are AI-decorated, not AI-native.

## Try the AI-native AI ERP
Demo runs in your browser. Install completes in 5 minutes. $0 forever after.
Try the demoInstall in 5 minutes

Related: read the architectural deep-dive at AI-native ERP, the books-only cut at AI accounting, the real estate view for property and lease books, the open-source map at open-source AI accounting, the migration playbooks at migrate, or the listicle 5 AI-native ERPs that earn the label.

---

## AI for NetSuite: Add an AI Analyst or Leave for Free
URL: https://www.erpclaw.ai/ai-for-netsuite/

# AI for your NetSuite business, whichever way you're leaning
Some businesses want AI beside NetSuite. Some want an AI-native system instead of it. ERPClaw covers both directions, and the second one is free forever. Here is each path, plainly.

## Keep NetSuite. Add an AI analyst beside it.
ERPClaw Insights answers plain questions about cash, customers and costs without touching your NetSuite setup. Connect your revenue sources, ask away, and your system of record stays exactly as it is. Automatic NetSuite sync is enabled per business: Insights reads your NetSuite data and posts the matching journal entries back, with your approval, so both stay current.
See Insights

## Leave NetSuite. Keep your history.
The full ERPClaw is an AI-native system: you run the whole business in plain language, and it costs nothing, forever, on your own server. Master records move over as spreadsheet exports, bank history comes straight from your bank, and the direct NetSuite importer, enabled per business, does it in one step.
How the move works

## Just comparing? Read the honest head-to-head.
What NetSuite does well, what it costs in practice, and where an AI-native system changes the daily experience, written plainly enough to forward to your accountant.
ERPClaw vs NetSuite

## What "AI-native" feels like day to day
NetSuite shops often carry a dedicated admin because the interface is the job: saved searches, custom forms, permissions. In an AI-native system, "show me June's profit by department" is the whole task. You ask; real accounting code does the work; the books get checked after every posting. No admin layer, because there is nothing to administer your way through.
That is the difference a rebuilt foundation makes, and it is why a wave of AI-native tools exists at all. The question is only whether you want that foundation beside your current system or under your whole business.

## Common questions

### What does 'AI-native' mean, in plain terms?
It means the plain-language request is the way you use the system, not a chat window bolted onto menus. You say 'invoice Acme for the March retainer, 4,500 dollars, net 30' and real accounting code posts a real, balanced entry. Systems designed before AI can add assistants, but the menus underneath remain the real interface.

### Does NetSuite have AI features?
NetSuite by Oracle has been adding AI features to its suite. The difference is architectural: those features assist you in operating NetSuite, while an AI-native system makes the conversation itself the way work gets done. Which one you want depends on whether NetSuite is staying.

### There are AI services that run on top of NetSuite. Is that the same thing?
Different animal. Tools like Billow put an AI accounting team on top of the NetSuite you keep paying for, which suits companies locked in for years. ERPClaw's answer is either an analyst beside NetSuite (Insights) or a free system to replace it. See the Billow head-to-head for the full picture.

### What does ERPClaw cost compared to NetSuite?
ERPClaw is $0 forever: open source, self-hosted, unlimited users. NetSuite pricing is negotiated and renewals climb. The trade is that you host ERPClaw yourself, on your own machine or server.

### Can I really move off NetSuite without a consulting engagement?
The file-based path is real and documented: export your lists as CSV, import them with explicit duplicate handling, bring bank history in from statement files, set opening balances. It is honest work, and it is yours rather than an invoice. The direct importer that does it in one step is enabled per business; request it below.

### What about multi-entity and consolidation?
Group structure, intercompany transactions and eliminations are real and tested in ERPClaw, and the eliminations are safe to re-run, with any leftovers from an earlier run named in the consolidated report rather than hidden. Currency translation for foreign subsidiaries is recorded as a deliberate entry rather than computed for you. Payments settle in the currency each invoice was raised in.

Related reading: ERPClaw vs Billow · Migrating from NetSuite · AI-native vs the AI layer

## We set this up for you, one business at a time.
Automatic NetSuite sync and the one-step importer are set up per business so they match your account. Leave your email and we set yours up.

---

## AI for QuickBooks: try AI-native instead
URL: https://www.erpclaw.ai/ai-for-quickbooks/

# Looking for AI in QuickBooks? Try AI-native instead.
Intuit Assist is a chat sidebar on top of a 1983 general ledger product. ERPClaw is AI-native: the agent posts journal entries autonomously, with constitutional invariants enforcing integrity. Open source, self-hosted, $0 forever. Free your data and your books.
Migration playbookTry the demo

## Why QuickBooks AI is not enough
Intuit Assist is real AI but it is decorated, not native. The underlying QuickBooks data model is the 1983 general ledger product with mutable history, a 5-user cap on Plus, and audit-trail concerns. Adding a chat sidebar does not fix those structural issues. ERPClaw rebuilds the architecture so the AI is the primary user; the data model is immutable, the user cap is gone, the audit trail is chain-hashed, and Stripe + Shopify are first-party.

- ·No user cap, no per-seat pricing.
- ·Immutable GL. Cancel = reverse, never edit.
- ·First-party Stripe + Shopify. No more manual fee reconciliation.
- ·$0 forever. Self-hosted on your hardware.
- ·Open source. Read the code; fork the project; own the future.

## QuickBooks Online + Intuit Assist vs ERPClaw
| Feature | QuickBooks + Intuit Assist | ERPClaw
| AI architecture | Intuit Assist (chat sidebar on 1983 GL) | AI-native by design from line one
| Cost | $38 (Simple Start) to $115 (Plus) per month | $0 forever
| User cap | Plus capped at 5 users; Advanced higher | No cap, self-hosted
| Audit trail | Mutable history (known concern) | Immutable GL by constitution
| Stripe sync | Manual fee reconciliation needed | Native integration with ASC 606
| Shopify sync | Reviewers report sync gaps | Native integration v1.1.3
| Self-hosting | No (Intuit cloud) | Yes (your hardware)
| Source code | Closed | Free & Open Source, github.com/avansaber/erpclaw
| Multi-currency | Premium tier required | 7 currencies, no paywall
| Inventory module | Plus tier; performance ceiling | FIFO/weighted-avg, multi-warehouse, BOM, work orders
| Payroll | Add-on subscription | Included module (US W-2, 1099, FICA, FUTA, SUTA)
| Industry verticals | None native | 14 native (healthcare, education, construction, etc.)

## Why teams move from QuickBooks to ERPClaw

### Intuit Assist is a chat sidebar
Intuit Assist is an AI helper that summarizes reports and suggests categorizations. The underlying QuickBooks data model is the 1983 general ledger product. The AI does not post journal entries on its own; a human still drives every meaningful write. That is the textbook definition of AI-decorated.

### QuickBooks audit trail has known concerns
Mutable-history accounting tools allow some financial entries to be edited without an obvious trail. ERPClaw enforces immutable GL by constitution; cancellations create reversal entries, never edits. A chain-hashed audit trail and an always-on checking engine make tampering structurally detectable.

### Manual Stripe reconciliation eats hours
QuickBooks does not natively understand Stripe processing fees, application fees, or refund splits. Users commonly report many hours per month manually adding fees and resolving amounts across transactions. ERPClaw ships a native Stripe integration with ASC 606 revenue recognition; daily auto-reconciliation removes the manual work.

### User cap is a structural ceiling
QuickBooks Online Plus caps at 5 users. The next plan is Advanced. Hitting the cap forces a tier upgrade for the whole team. ERPClaw has no user cap, no seat-based pricing, no tier you outgrow. Self-hosted on your hardware, every user is free.

### Performance ceiling at SMB scale
QuickBooks data files have a structural ceiling. Performance can degrade as transaction count, item count, or user load grows past common SMB thresholds. ERPClaw uses SQLite with WAL mode and indexed queries; PostgreSQL takes over for genuinely large workloads with no schema change.

### Lock-in compounds across the ecosystem
Every QuickBooks add-on (Synder, A2X, Bookkeep, Webgility) is a separate SaaS subscription on top of QBO. The combined annual cost is significant for a small business. ERPClaw replaces the entire stack with one self-hosted system at $0 forever.

## QuickBooks Online or QuickBooks Desktop? The migration is different.
Both paths land in the same place (ERPClaw, AI-native, $0). The export step is what differs. Pick the path that matches your current setup.

### From QuickBooks Online (QBO)
QBO has CSV exports for chart of accounts, customers, vendors, items, and journal entries directly from the report center. Run the exports, install ERPClaw, run the import actions, parallel-run for a week, then cut over. Full procedure at /migrate/from-quickbooks/. Plan on a weekend for a small business; longer for multi-entity setups.

### From QuickBooks Desktop
Desktop exports go through IIF or Excel from the company file. Same import actions on the ERPClaw side; the wrinkle is field-naming differences between Desktop and Online CSV shapes. For Desktop migrations, talk to a co-founder via /demo/ first; we'll size the export-to-import mapping with you so the migration doesn't surprise on day three.

Whether you're on QBO or Desktop, the connected modules carry over. AI bookkeeping handles the daily volume; AI inventory handles items, BOMs, and multi-warehouse stock; AI accounting covers the broader category. The architectural argument for picking open source is at open-source AI accounting; the head-to-head comparisons against mid-market peers are at NetSuite, Sage Intacct, and Rillet.

## Frequently asked questions

### Does QuickBooks have AI?
Yes, QuickBooks ships Intuit Assist, a chat sidebar that helps with categorization, report summaries, and Q&A about your books. It is AI-decorated, not AI-native: the underlying data model is the 1983 general ledger product, and the AI is a feature inside an existing product. The agent does not post journal entries autonomously the way ERPClaw's AI-native architecture does.

### Is there an AI-native alternative to QuickBooks?
Yes, ERPClaw. It is AI-native by architecture: every action across the platform (post-gl-entries, add-customer, reconcile-stripe-payout, and many more) is invokable by an AI agent without UI clicks. The agent posts journal entries autonomously with constitutional invariants enforcing data integrity. ERPClaw is also Free & Open Source, self-hosted, and $0 forever.

### Can I switch from QuickBooks to ERPClaw without losing data?
Yes. The QuickBooks-to-ERPClaw migration playbook at /migrate/from-quickbooks/ is a 7-step weekend procedure. Export from QBO (trial balance, customer/vendor/item lists, chart of accounts, journal entries). Install ERPClaw. AI agent reads the export and runs ERPClaw's import actions. Spot-check trial balance and aging. Run parallel for one week. Cut over.

### Is ERPClaw really comparable to QuickBooks for a small business?
Yes for most small businesses. ERPClaw covers accounting (immutable double-entry GL), invoicing, payments, payroll, inventory, and US tax forms (W-2, 1099, NACHA, FICA, FUTA, SUTA). What it does not have today is a polished web dashboard for non-technical users; the primary interface is an AI agent in a chat. If your team will adopt chat-first, ERPClaw is a structural upgrade.

### Will my CPA accept books from an open-source AI tool?
Yes. ERPClaw exports trial balance, P&L, balance sheet, AR aging, AP aging, GL detail, and the core financial report set as CSV or PDF in standard formats. The chain-hashed audit trail and immutable GL produce books that pass external review cleanly. CPAs increasingly prefer this style of structured, auditable books over mutable-history alternatives.

### What does ERPClaw cost compared to QuickBooks?
ERPClaw is $0 forever (open source, self-hosted). QuickBooks Online ranges from $38 per month (Simple Start) to $115 per month (Plus, capped at 5 users) to $235 per month (Advanced), per Intuit's pricing page, checked July 2026. Compounded over 5 years, the cost difference is significant for a small business. The architectural advantages (open source, AI-native, no vendor lock-in, no user cap) are upside on top of the cash math.

### Can ERPClaw replace QuickBooks plus all my add-ons?
Yes for most setups. ERPClaw includes native Stripe (with ASC 606), Shopify (v1.1.3), inventory, payroll, and accounting. The QuickBooks plus Synder plus A2X plus Webgility plus Bookkeep stack collapses to one self-hosted ERPClaw install. One database, one audit trail, one bill (which is $0 because there is no bill).

Related: read the QuickBooks comparison, the QuickBooks import path, the migration playbook, the AI accounting story, the AI bookkeeping guide, or the long-form QuickBooks alternative post.

---

## AI inventory management, open source and free
URL: https://www.erpclaw.ai/ai-inventory/

# AI inventory connected to your books, not bolted on
Cin7, Katana, and Inflow added forecasting AI to a forms-and-workflows inventory product. ERPClaw is AI-native: the agent posts stock-ledger entries directly with FIFO or weighted-average costing, in the same transaction as the GL entry. Open source, self-hosted, $0 forever.
Try the demoInstall in 5 minutes

## What is AI inventory management?
AI inventory management is inventory software where an AI agent reads stock levels, posts movements, computes costs, and proposes actions. AI-native inventory has the agent at the database layer with constitutional invariants enforcing inventory plus GL parity. AI-decorated inventory (most of the market) bolts a forecasting model onto a 1990s ERP. ERPClaw is AI-native and connects inventory to your books in a single shared database.

- ·SLE plus GL in one transaction. No nightly sync drift, no reconciliation queues.
- ·FIFO or weighted-average costing per item per warehouse.
- ·Multi-warehouse with reservations. Sales orders, work orders, transfers all share authoritative stock.
- ·BOM and work orders. Manufacturing flow with WIP costing.
- ·$0 forever, self-hosted. SQLite default; PostgreSQL fully supported.

## Ask it like you'd ask a person
Stock questions answered the way you would ask a colleague, from live quantities.
"How many of these can I still promise customers?"
On hand minus what is already reserved, which is the only honest answer to that question. What the system says you can sell is what you can actually sell.

"Hold 40 of the blue chairs for this order"
A hard hold, refused up front if the stock is not there, so a promise never gets quietly broken at picking time.

## What AI inventory does in ERPClaw

### Stock movements posted by an AI agent
When a Shopify order ships, the AI posts the stock-ledger entry (SLE) and the GL entry in the same transaction. Inventory and accounting move in lockstep. No nightly batch sync, no reconciliation drift.

### FIFO and weighted-average costing
Both costing methods are first-class. Pick per-warehouse or per-item. The agent computes cost-of-goods-sold using the correct method on every shipment with full audit trail.

### Reorder points and stock alerts
Set reorder points on items; the agent monitors stock levels and creates purchase orders or alerts your team when inventory falls below threshold. No separate forecasting SaaS subscription.

### Multi-warehouse stock with reservations
Multiple warehouses, transfer orders between them, reservations against sales orders. The shared database means a sales order in one warehouse and a manufacturing work order in another see the same authoritative stock level.

### Bill of materials and work orders
BOM definition with sub-assemblies, work orders that consume raw materials and produce finished goods. The agent posts work-in-progress and finished-goods movements correctly for accountants and operators.

### Cycle counts and physical inventory
The agent runs cycle counts on a schedule, surfaces variances, and posts adjustments with reason codes. Physical inventory at year-end reconciles against the perpetual SLE without separate reconciliation tooling.

## AI inventory vendor comparison
| Vendor | Focus | AI posture | License | Cost
| Cin7 | Inventory + order management for omnichannel retail | Forecasting and demand planning AI add-on | Proprietary | $349-$1,499+/month
| Katana | Manufacturing-focused inventory + production planning | AI assist for production scheduling | Proprietary | $179-$899/month
| Inflow | SMB inventory with light manufacturing | Limited (rule-based reorder) | Proprietary | $110-$549/month
| NetSuite SuiteAnalytics | Enterprise inventory inside NetSuite | AI add-on (NetSuite AI) | Proprietary | $30K+/year (part of NetSuite)
| ERPClaw | Full ERP inventory connected to GL with FIFO/weighted-avg | AI-native (agent posts SLE entries directly) | Free & Open Source | $0 forever

## Who runs AI inventory on ERPClaw
Three operator profiles where AI-native inventory connected to the GL beats forecasting-only SaaS. The connecting thread: stock and books in one shared database with one audit trail.

### Shopify and ecommerce operators
Shopify orders ship, COGS posts in the same transaction as the stock-ledger entry, and the bookkeeping side stays in lockstep. No nightly batch sync. First-party Shopify integration at /docs/shopify/. Connected bookkeeping deep-dive at /ai-bookkeeping/.

### SMB manufacturers
Items, BOMs with sub-assemblies, work orders that consume raw materials and produce finished goods. The agent posts WIP and finished-goods movements correctly for both operators and accountants. NetSuite-level depth without the NetSuite price; the side-by-side detail at ERPClaw vs NetSuite.

### Multi-warehouse distributors
Multiple warehouses with their own FIFO or weighted-average bases. Transfer orders post the right cost on each side. The shared database means a sales order in warehouse A and a work order in warehouse B both see authoritative stock. For QuickBooks teams running inventory across locations, the structural upgrade is at AI for QuickBooks and vs QuickBooks.

For the broader category framing see AI accounting and the architecture argument at open-source AI accounting. For close-specific finance comparisons see vs Sage Intacct and vs Rillet.

## Frequently asked questions about AI inventory

### What is AI inventory management?
AI inventory management is software where an AI agent reads stock levels, posts movements, computes costs, and proposes reorder actions on its own. AI-native inventory (like ERPClaw) integrates the agent at the database layer; AI-decorated inventory (Cin7, Katana, Inflow with AI add-ons) is a forecasting layer bolted onto a forms-and-workflows product.

### Is forecasting AI the same as AI-native inventory?
No. Forecasting AI predicts demand; the human still posts stock movements through forms and approvals. AI-native inventory has the agent posting movements directly into the ledger with constitutional invariants enforcing inventory plus GL parity. Forecasting AI is a useful add-on; AI-native is a different architecture.

### Can ERPClaw handle multi-channel retail (Shopify, Amazon, eBay)?
Today ERPClaw's first-party connectors are Shopify and Stripe. Amazon and eBay can be ingested via CSV or through a thin custom action defined in a SKILL.md spec. We are prioritizing first-party connectors based on user demand. For Shopify-only stores, ERPClaw covers the stack natively.

### How does ERPClaw inventory compare to NetSuite for manufacturing?
NetSuite is broader at the Fortune 500 level (advanced manufacturing, work-in-progress costing, complex BOMs with sub-assemblies of sub-assemblies). ERPClaw covers SMB to mid-market manufacturing: items, warehouses, BOMs, work orders, FIFO/weighted-average costing, cycle counts, reorder points. The architectural advantage at SMB to mid-market is the AI-native action layer plus $0 forever cost.

### Does ERPClaw support multi-warehouse with FIFO costing per warehouse?
Yes. Each warehouse has its own SLE; FIFO and weighted-average costing are configurable per item per warehouse. Transfer orders between warehouses post the right cost basis on each side. The shared database means reports across warehouses are accurate without batch sync.

### Can the AI agent close my month-end inventory reconciliation?
Yes. The agent runs the close calendar: physical-vs-perpetual variance, work-in-progress reconciliation, COGS recalculation, period-end accruals. You review exceptions, not entries. The invariant checks ensure inventory and GL stay balanced; structural drift is impossible.

### Is open-source AI inventory really production-ready?
ERPClaw inventory is on the same code path as the accounting ERP serving paying Stripe Marketplace customers. Extensive automated test coverage, immutable stock-ledger entries (cancel = reverse, never edit). The open-source advantage is observable controls; you can read the code and verify the costing logic line by line.

Related: read the broader AI ERP guide, the AI-native ERP framework, the foundation features for chart of accounts and GL, the manufacturer page, the restaurant view for food cost and waste, or the FIFO vs weighted-average deep-dive.

---

## AI-Native ERP: Definition, 5-Trait Test, and Buyer Checklist
URL: https://www.erpclaw.ai/ai-native-erp/

# AI-native ERP: the 5-trait test that exposes AI-decorated bolt-ons
Most "AI-native ERP" claims are AI-decorated bolt-ons on 1990s forms-and-workflow systems. SAP, Oracle, NetSuite, ERPNext, and Odoo all fail the architectural test. Here's the test, the three-tier map, and the only open-source AI-native option.
By Nikhil Jathar, Co-founder, ERPClaw · 2026-05-05 · ~14 min read

## What is AI-native ERP?
AI-native ERP is enterprise software built around AI agents that reason, recommend, and act on the data, not bolted-on copilots that only summarize a forms-and-workflows ERP. The 5-trait test below distinguishes the two: user experience, workflow, data, automation, and governance must each be agent-first. By that test SAP Joule, Oracle AI Agents, NetSuite, Odoo, and ERPNext fail; ERPClaw is the open-source AI-native option.

- ·UX is conversational: the primary interface is chat with role-based, proactive AI assistants, not menus and forms.
- ·Workflows are agentic: agents reason and act, not rule-based approvals plus batch jobs.
- ·Data is semantic: contextual, explainable business knowledge, not just tables and BI dashboards.
- ·Automation is embedded: AI runs inside finance, supply chain, HR, and procurement; not external RPA scripts.
- ·Governance covers AI: traceability, human approval gates, and model governance, not just role security and audit logs.
Next: evaluate the category, inspect ERPClaw capabilities, or try a workflow.

AI-native ERP is the most-claimed term in enterprise software in 2026. Most vendors who claim it are running AI-decorated bolt-ons on top of forms-and-workflow ERPs designed in the 1990s and 2000s. SAP SE's Joule is an "AI assistant" by SAP's own marketing. Oracle Corporation calls theirs "AI Agents." Microsoft Corporation positions Dynamics 365 with "Copilot." None of them claims AI-native architecture, because none of them has it. This article is the test that exposes the gap.
This article is the answer. We will walk through:

- ·A testable 5-trait definition (borrowed from how ChatGPT itself maps the category)
- ·The three-tier map: enterprise AI-enhanced, commercial AI-native startups, open-source AI-native
- ·Where ERPClaw fits and how it scores on each of the 5 traits
- ·A 13-question evaluation checklist for any vendor claiming AI-native
- ·What ERPClaw doesn't do yet (honest)
If you're a CTO, VP Operations, founder, or finance lead trying to evaluate the category, this is the framework. If you have already decided you want open-source AI-native and want the practical map, jump to open-source AI accounting.

## Why this matters in 2026

### The 30-year ERP pattern
Every ERP since SAP R/3 in 1992 follows the same recipe. Forms capture data. Workflows route approvals. Batch jobs reconcile and post. BI dashboards summarize the result. NetSuite, Microsoft Dynamics, Sage Intacct, Oracle Fusion, and ERPNext are all variations on this theme. The recipe is mature, well-understood, and battle-tested. It also assumes a human is driving every meaningful write to the database.

### The shift Microsoft is naming
Microsoft Corporation has started talking about ERP as a shift from system of record to system of action. The framing also surfaces in how ChatGPT itself maps the category. The idea is simple: if AI agents can interpret signals and execute against business systems, the ERP stops being a passive ledger that humans update and starts being an active participant. Records still matter; the writer changes.

### What this means for a buyer in 2026
For a buyer evaluating ERP this year, the question is no longer "does this vendor have an AI roadmap." Every vendor does. The question is whether the system is built so an AI agent can drive it safely. That is a question about architecture, not features. An AI assistant that suggests a journal entry for a human to approve is one architecture. An AI agent that submits the journal entry through a validated action layer with an immutable audit trail is a different architecture. The 5-trait test below is how you tell them apart.

## The 5-trait test
The cleanest framing of what AI-native means came from asking ChatGPT to map the category. We are quoting it verbatim because it captures the right five layers in one paragraph.
"An AI-native ERP should have five core traits across user experience, workflow, data, automation, and governance."
| Layer | Traditional / AI-decorated | AI-native
| User experience | Menus, forms, reports | Conversational, role-based, proactive assistants
| Workflow | Rule-based approvals + batch jobs | Agentic workflows that reason, recommend, and act
| Data | Tables, reports, BI dashboards | Semantic, contextual, explainable business knowledge
| Automation | RPA, scripts, scheduled jobs | AI agents embedded into finance, supply chain, HR, sales, procurement
| Governance | Role security + audit logs | Role security + AI controls, human approval, traceability, model governance

### Layer 1: user experience
AI-decorated UX is menus, forms, and reports with a chat sidebar that pre-fills fields. ERPNext plus a chatbot widget is the textbook example; NetSuite plus a Bill.com plug-in is the same shape at enterprise scale. AI-native UX puts the conversation in the centre. Rillet's conversational close, Numeric Inc.'s reconciliation flow, and ERPClaw's chat-first action layer all start from "what does the user want to do" and let the AI map that to actions, not "what menu item does the user click."

### Layer 2: workflow
Rule-based workflow is a 10-step approval matrix in NetSuite. Every condition is hard-coded; every exception is a change request. Agentic workflow is "the AI flags this PO for human review because the supplier flagged late delivery three times in the last 90 days, and here's the recommendation." The reasoning is explicit, the data behind it is queryable, and the human stays in the loop on judgment calls while the routine flagging runs itself.

### Layer 3: data
A BI dashboard is a quarterly P&L pivot table. It tells you what happened. A semantic data layer is "the AI knows revenue is recognized over 12 months for SaaS contracts and tells you when it sees a contract that doesn't match policy." The schema is exposed; the AI can introspect; the explanations cite the rows and rules behind any answer. You can ask "why" and get the lineage, not just the number.

### Layer 4: automation
RPA is a bot that fills a form. It breaks when the form changes. An AI agent is "given a new vendor invoice, the agent reads OCR, matches to PO, flags 3-way variance, recommends approval or hold." ERPClaw's `match-vendor-invoice` action is the worked example: the AI invokes the action, the action runs the matching logic, and the result is either a clean post or a flagged exception with reasoning attached.

### Layer 5: governance
Role security plus audit logs is the current standard. AI controls add a second loop: every AI action is gated on a per-invocation user-confirmed flag for state-mutating operations, the immutable audit log records every AI invocation with before-and-after state, and the trust root for any auto-update is cryptographically signed. The AI can read freely; it cannot write to the GL without the gate, and it cannot bypass the gate by setting an environment variable.

## The three-tier map
Once you have the 5-trait test, the category map sorts itself. There are three tiers with different architectures and different best-fit buyers.
| Tier | Players | Architecture | Strengths | Honest limits
| Enterprise tier (AI-enhanced) | SAP Joule, Oracle AI Agents (Fusion Applications), Microsoft Dynamics 365 Copilot | Mature ERP + AI assistant layer | Multi-entity, deep modules, enterprise governance, decades of ERP depth | AI is added on top; architecture is forms-and-workflows; expensive
| Commercial AI-native startup | Rillet, Numeric, Puzzle, Digits, Campfire | AI-native by design, cloud SaaS | Fast close, AI-native UX, modern stack | Closed source; vendor SaaS; finance-only scope (not full ERP)
| Open-source AI-native | ERPClaw | AI-native by design, open source, self-host, full ERP | Open + free + self-host + full ERP scope | Chat-first today (web dashboard included, beta); SMB-focused (not Fortune 500)

### Enterprise tier: AI-enhanced
SAP SE calls Joule "an enterprise AI assistant." Oracle Corporation calls theirs "AI Agents for Fusion Applications." Microsoft Corporation positions Dynamics 365 with "Copilot" inside it. None of them claims AI-native architecture, because none of them has it. SAP S/4HANA, Oracle Fusion, NetSuite, and Dynamics 365 are forms-and-workflows ERPs from the 1990s and 2000s with a 2024 to 2026 AI assistant layered on top. That is a defensible architectural choice with mature underlying scaffolding, multi-entity depth, and decades of enterprise governance behind it. It is a different category from systems built with AI as the action layer from the floor up. Their AI is real and well-funded; the architectural reality is that it is AI-decorated by the vendors' own marketing language.

### Commercial AI-native startup tier
The 2024 to 2026 cohort: Rillet, Inc., Numeric Inc., Puzzle, Digits, and Campfire. Cloud SaaS, finance-close-focused, AI-native by design from commit one. Best fit for VC-backed startups and SaaS finance teams who want polished close-specific UX, vendor-managed updates, and a finished product on day one. Closed source and vendor SaaS are real trade-offs: your books live on the vendor's server, your data residency is the vendor's call, and the AI prompts and policies are the vendor's IP. For teams with a finance-only scope and budget for a per-seat subscription, this tier is often the right answer. For the head-to-head with the most-cited startup in the cohort, see ERPClaw vs Rillet.

### Open-source AI-native (single entry: ERPClaw)
ERPClaw sits in a category by itself today: open source, self-hosted, full ERP scope (CRM, AR, AP, GL, payroll, tax, inventory, integrations, 14 industry verticals), with patent pending and trademark filed. The first open-source AI-native ERP we know of, in the sense that we couldn't find a peer that's both fully open source AND AI-native by architecture rather than by add-on plug-in. ERPNext is open source but its AI is a plug-in. Odoo Community is open source but Odoo Enterprise gates the AI features. We are honest about being CLI-first today (web dashboard in progress for Stripe and Shopify) and SMB-focused rather than Fortune 500. For the practical map of the open-source corner, see open-source AI accounting.

## ERPClaw on the 5-trait test
Show, don't tell. Here is how ERPClaw scores on each of the five layers, with concrete examples that you can reproduce on your own machine after a one-line install.

### Trait 1: user experience
Today, ERPClaw runs from chat (`openclaw chat`), CLI (`erpclaw <action>`), and Telegram (the OpenClaw test bot). A web dashboard is live for some flows and expanding. Concrete example: type "set up a company called Acme Imports" and the AI invokes the `setup-company` action; 94 GAAP-aligned accounts seed in one transaction; you can immediately invoice a customer or pay a contractor. No menu hunting, no form, no copy-paste from a chart-of-accounts template.

### Trait 2: workflow
The action layer IS the API. Every action in the system is a single SQLite or PostgreSQL transaction with full rollback on failure. Money is stored as Decimal-in-TEXT, the GL is immutable (cancel equals reverse), and every posting runs through a 12-step validation pipeline before it touches the books. Concrete: the AI sees "submit-payment for invoice 12345"; the router enforces the per-invocation user-confirmed flag; if it is missing, the action returns a clean error and nothing posts. The AI cannot violate the GL even when it is wrong about something else.

### Trait 3: data
One shared database (SQLite by default, PostgreSQL via PyPika as a first-class alternative). One schema across every module, with WAL mode, foreign-key enforcement, and the audit log immutable by schema (no `updated_at` column). The schema is exposed; the AI introspects table definitions when it needs to answer questions like "what's the running balance on the Stripe clearing account today" without anyone hand-writing SQL. Open source means you can read every CREATE TABLE and every constraint.

### Trait 4: automation
The module set covers the operational footprint of an SMB: full foundation plus payroll, payments, inventory, manufacturing, compliance, integrations, and 14 industry verticals (retail, food, hospitality, legal, healthcare, education, property, automotive, agriculture, construction, nonprofit, plus four regional packs). The Stripe addon (every operation, live on the Stripe Marketplace) handles real-world e-commerce automation including ASC 606 deferred revenue and net-of-fees payout decomposition. The Shopify addon (every operation, v1.1.3) mirrors order, fulfillment, and refund state into the books. The AI invokes any of them by name from natural language.

### Trait 5: governance
v4.1.x runtime gate: state-mutating actions require a per-invocation user-confirmed flag the AI cannot bypass through environment variables. Immutable audit log records every action with full input, full output, and the exact GL entries that posted. The module registry is ed25519-signed for foundation reconciliation; the trust root fingerprint is `d471:335b:0e4d:75ce`; the strict-mode loader refuses unsigned, tampered, or downgraded registries. Open source under Free & Open Source means you can fork the governance layer and audit it line by line.

## 13-question evaluation checklist
Use this with any vendor claiming AI-native, including ERPClaw. Each question has a "what good looks like" rubric. Some questions ERPClaw answers strongly; some it does not. The checklist is reusable for any vendor evaluation.

### Architecture and AI-nativeness

- Is the action layer the API, or is the AI a chat box on top of forms?
What good looks like: Every business action invokable from a prompt with no UI dependency. The same action runs from chat, CLI, or a web button.

- Can I invoke any business operation from natural language?
What good looks like: Type "add Bob from BigCo as a customer" and it lands in one transaction with the right defaults, not a form pre-fill that still needs a click.

- Does the system record every AI invocation in an immutable log with before-and-after state?
What good looks like: Per-action audit row, no UPDATE on the audit table, cancel equals reverse. You can replay any AI decision later.

- Does it gate state-mutating AI actions with explicit user confirmation?
What good looks like: Dangerous actions need an explicit per-invocation flag the AI cannot bypass. No silent environment-variable shortcut.

- Vendor-locked or model-agnostic?
What good looks like: You can swap GPT-5 for Claude or a local Ollama model without re-platforming. Model choice belongs to the buyer, not the vendor.

### Hosting and economics

- Self-host or vendor SaaS, and where does the data live?
What good looks like: Data on hardware you control, or vendor SaaS that meets your residency and compliance bar. The honest answer beats marketing copy.

- Open source or closed, and what license?
What good looks like: Free &amp; Open Source or Apache 2.0 lets you fork. AGPL is viral. Closed source means you cannot audit the GL math or the AI prompts.

- Database backend and migration path?
What good looks like: At least two backends supported (for example SQLite plus PostgreSQL) with documented migration. Vendor cloud DB only is a lock-in red flag.

### Scope and integrations

- Finance-only or full ERP?
What good looks like: Matches your operational scope. Broader is not always better; finance-only is the right choice for many SaaS startups.

- Native Stripe, Shopify, and bank, or plug-in stack?
What good looks like: Native means vendor-supported with a roadmap. Plug-in means community-maintained and brittle on every platform update.

### Honest gap-checking

- Multi-entity and multi-currency depth?
What good looks like: Depends on your operations. NetSuite and SAP S/4HANA win on intercompany consolidation, FX revaluation, and multi-jurisdictional tax. AI-native startups and ERPClaw are weaker here today.

- Polished close-cockpit UX?
What good looks like: Depends on your team. Rillet and Numeric have dedicated reconciliation queues, variance review, and board-ready P&L flows built as one screen. ERPClaw closes the books inside the ERP that runs everything else; the difference is the shape of the screen, not the capability.

- Vendor support and roadmap accountability?
What good looks like: Depends on your org. SaaS vendor with paid support tiers is a feature for some teams. Open source plus co-founder access is a feature for others. In-house IT is a third valid model.

## What ERPClaw doesn't do yet
Trust signal. We name the gaps so you can decide whether they matter for your operation.
Status as of August 2026: the web dashboard is included free and currently in beta, building its screens for every installed module automatically, while chat and CLI remain the primary way to run state-mutating actions. The Stripe addon is selling on the Stripe Marketplace; the Shopify App Store listing is still pending Partners-dashboard work. The list below reflects that current state.

- ·No continuous online learning. AI-native for us means the action layer is the API, not a self-improving model. We're rules plus LLM plus actions, not AGI.
- ·Web UI for Stripe and Shopify is still CLI-first. Chat works; the polished web dashboard is in progress.
- ·Shopify App Store listing pending. Stripe Marketplace IS live and selling; Shopify listing waits on Partners-dashboard work.
- ·Deepest in the US. Regional modules cover Canada, EU, India, and UK alongside US in core. English-only docs today.
- ·Multi-currency Phase 1. Invoice currency must equal payment currency; no FX gain or loss in our books. That keeps the GL clean. ASC 830 FX revaluation is intentionally out of scope, not a roadmap commitment.
- ·Not Fortune 500 scale. Designed for SMB and growth-stage. NetSuite and SAP S/4HANA serve a different segment with different governance and consolidation requirements.

## Frequently asked questions

### What's the difference between AI-native and AI-decorated ERP?
AI-decorated means the underlying ERP (forms, workflows, tables, reports) was designed before AI existed and a chatbot or copilot was bolted on later. AI-native means the action layer is the API; every business action is invokable from a natural-language prompt; the AI does not translate-then-form-fill, it directly invokes the action with an audit trail of before-and-after state.

### Is "agentic ERP" the same as "AI-native ERP"?
Roughly yes. Agentic emphasizes that the AI initiates actions instead of only answering questions. AI-native emphasizes the architecture is built around AI from the start. Microsoft uses both terms. The meaningful distinction is form-bolt-on versus action-as-API; if the AI can submit a state-mutating business action with full audit trail, the label matters less than the architecture.

### Why isn't SAP Joule considered AI-native?
Joule is real, well-funded enterprise AI. The architectural question is whether the underlying ERP changed. It did not; SAP S/4HANA is a forms-and-workflows ERP. Joule is the AI assistant layered on top. SAP itself positions Joule as an "AI assistant embedded across SAP and non-SAP systems," not as a re-architecture. That is a defensible choice with mature scaffolding underneath, but it is a different category from systems built with AI as the action layer.

### Can ERPClaw replace NetSuite for a 100-person company?
For most US SMBs scaling toward 100 people, yes. ERPClaw covers AR, AP, GL, payroll (W-2, 1099, NACHA, FICA, FUTA, SUTA), inventory, multi-entity, intercompany, and consolidation. The honest gap: NetSuite has more polished multi-currency, deeper international payroll, and 25 years of enterprise governance scaffolding. If your operations are US-first and SMB-scale, ERPClaw fits. If you have 12 entities across 8 currencies, NetSuite still wins.

### Is ERPClaw the first open-source AI-native ERP?
Yes. As of 2026 there is no peer that's both fully open source AND AI-native by architecture. ERPNext is open source but its AI is a plug-in stack on top of a forms-based core. Odoo Community is open source but Odoo Enterprise gates the AI features. Akaunting and Manager.io don't have AI. ERPClaw is the first open-source AI-native ERP. If a peer surfaces, we will update this page; until then, the category has one entry.

### How do I evaluate whether a vendor's AI claim is real?
Use the 13-question checklist on this page. The fastest single test: ask the vendor to show a state-mutating business action invoked entirely from natural language with an audit log of before-and-after state. If they show forms with AI pre-fill, it's AI-decorated. If the AI submits the action and the audit log records the invocation, it's AI-native.

Looking specifically at AI for the books, not the full ERP? See the AI accounting pillar.

## Where to go next
Five paths from here, depending on what you want to do next.
Try the demo
Choose a role and follow a guided business conversation.
Install in 5 minutes
`clawhub install erpclaw` · full docs
Open-source map
If open source plus self-host is the deciding factor.
ERPClaw vs Rillet
Head-to-head with the most-cited AI-native peer.
Compare AI ERP options
The buyer's guide cut: vendor matrix, costs, FAQs.
github.com/avansaber/erpclaw
Read the source. open source. 230,000+ lines of audited Python.

Long-form: read the 5 AI-native ERPs that earn the label, AI-decorated vs AI-native software, the Odoo alternative built AI-native from day one, or ERPNext vs ERPClaw. The data behind the 5-trait test lives in the research hub, including the AI-native vs decorated five-trait study.

## Sources

- ChatGPT GPT-5, conversation captured 2026-05-05 (5-trait framing and three-tier map)
- SAP SE Joule product pages (sap.com/products/artificial-intelligence/ai-assistant.html)
- Oracle Corporation, AI Agents for Fusion Applications (oracle.com/applications/cloud/ai-agents/)
- Microsoft Corporation, Dynamics 365 agentic AI pages (microsoft.com/dynamics-365)
- Rillet, Inc. (rillet.com) and Numeric Inc. (numeric.io)
- ERPClaw `module_registry.json` (registry_version 8, the module registry that drives the whole system)
- ERPClaw on the Stripe Marketplace

---

## Audit-Ready Books: Provably Correct Accounting
URL: https://www.erpclaw.ai/audit-ready-books/

# Audit-ready books, without doing anything extra
Most software makes clean books your job. ERPClaw, the Free & Open Source ERP you host yourself, makes them the system's job: every entry is checked the moment it posts, mistakes are corrected in the open, and the whole history stays explainable. When the auditor, the bank, or the buyer shows up, you are already ready, and it cost you $0.
Get ERPClaw freeSee it in action

## How the books stay provable
Four things happen automatically, on every entry, that most systems leave to discipline and luck.

### Every entry is checked before it stands
A checking engine runs over the books after every single posting. An entry that would break the accounting, debits that do not match credits, money appearing from nowhere, is refused on the spot instead of discovered at year end.

### Nothing is ever quietly edited
Posted entries are permanent. A mistake is fixed with a reversal that shows what changed and when, the way an auditor expects, so the history always tells the whole story. There is no delete key for the past.

### Every action leaves a trail
Who asked for what, what the system did, and what it posted: recorded, every time, automatically. When someone asks 'why is this entry here', the answer is a lookup, not an investigation.

### The engine itself can be inspected
ERPClaw is fully open source. Your accountant, your auditor, or your own developer can read the exact code that keeps the books, which is a level of trust no closed system can offer at any price.

"An audit is only frightening when the books have secrets. We built ERPClaw so they cannot keep any: every entry is checked the moment it posts, every correction happens in the open, and every question has a lookup instead of an investigation. A boring audit is the product."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## Two questions you can always answer
"Are the books actually balanced?"
One question, answered by real double-entry math on the spot, not a dashboard's opinion. If they ever were not, the entry that broke them would have been refused when it was made.

"Why is this entry in the books?"
The trail is a lookup: who asked for it, what the system did, what it posted, and when. The ledger itself is permanent, so the answer cannot have been edited after the fact.

## And they stay provable through upgrades
Keeping books provable on the day you post an entry is the easy half. Keeping them provable through every upgrade, repair, and correction afterwards is the half that decides whether you can still trust the history a year from now.

### A release has to prove itself against the books
Before a version ships, it has to run complete business scenarios from beginning to end, driven in plain business language, and a deterministic check reads the books that come out and decides pass or fail. A scenario that ends with the wrong balance stops the release.

### Upgrades say what they touched
When an upgrade repairs something in your history, it records an audit-trail row for every document it changed, inside the same transaction as the change. Nothing about your past gets quietly tidied.

### Corrections are visible and confirmed
Group consolidation is safe to run twice, and if an earlier repeat run left damage behind, the consolidated trial balance names the leftover rows instead of absorbing them. Removing them takes an explicit confirmation, and every removal is recorded.

### A closed route beats a quietly wrong one
Shortcuts that could move stock quantity, or post group adjustments, without a matching entry in the books were retired rather than patched. They now decline and name the flow that does the job properly.

These arrived in ERPClaw 4.15, written up intested the way you actually work.

## The days this pays for itself

### The loan application
The bank wants financials it can trust. Yours come out of a system that provably balances, with the trail to show it.

### Due diligence
A buyer or investor sends their accountants in. Clean, permanent, explainable books are the difference between a week and a quarter.

### Tax season
Your accountant gets books that reconcile instead of a shoebox with software around it. Their hours drop, and so does their bill.

### The bad-surprise day
When a number looks wrong, the trail shows exactly what happened and when. No archaeology, no blame roulette.

Related: three-way matching before money leaves · profit by department or project · work with your books from your AI tools · how ERPClaw is tested

## Common questions

### What does 'audit-ready books' actually mean?
It means the books are kept in a way an auditor can verify: every entry balanced and checked at the moment it was made, corrections done as visible reversals rather than silent edits, and a complete trail of who did what. In ERPClaw that is how the system works by construction, not a report you generate at the end.

### Does the AI make this riskier?
The opposite, because of how ERPClaw is built. The AI decides what you meant; deterministic accounting code decides what gets posted; and the checking engine verifies the books after every posting. The AI never has a pen of its own.

### Can I show this to my accountant?
Please do. The short version for them: double-entry ledger, immutable postings with reversal-based corrections, full audit trail, and open-source posting logic they can read. Most accountants relax visibly at the second item.

### Is this an extra module or a paid tier?
Neither. It is how the free product keeps books. ERPClaw is $0 forever, self-hosted, open source, and the checking engine is not optional, for you or for us.

### What happens to my books when I upgrade?
Upgrades are held to the same standard as the books themselves. A version has to clear complete business scenarios judged against the resulting books before it ships, and any upgrade step that repairs existing data records what it changed, document by document, in the same transaction as the change. If a repair happens in your history, you can read exactly what it was.

### Does this replace an actual audit?
No, and we will not pretend it does. Audits are performed by auditors. What audit-ready books change is what the audit costs you in time, fees, and surprises, because the evidence an auditor needs already exists in order.

### What if I need accounting help, not just software?
We make software, and we offer implementation help getting set up. For bookkeeping, tax, or close work done by people, our accounting partners can help; reach out through the support page and we will connect you.

## Books you never have to apologize for
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawInstall guide

---

## Blog - AI-Native ERP Insights, Stripe + Shopify Updates
URL: https://www.erpclaw.ai/blog/

# Blog
Engineering deep-dives, product launches, and industry analysis. Stripe ASC 606, Shopify payout reconciliation, and the AI-native ERP shift.

VisionAugust 18, 2026

## Model Independence, With the Receipt
ERPClaw has run in production through four different models with no module changed. Why that is possible, why most AI products cannot claim it, and what it tells a buyer.
Read post

EngineeringAugust 16, 2026

## One Writer Per Table: Write Ownership in an ERP
Read-many, write-one as an architecture rule: any module may read any table, but only the owning module may write to it. Why that makes bugs findable.
Read post

LaunchAugust 15, 2026

## ERPClaw 4.15: Tested the Way You Actually Work
ERPClaw 4.15 is live: every release now clears complete business scenarios checked against the real books, and short payments are written off in the payment.
Read post

EducationAugust 13, 2026

## AI-Native vs the AI Layer: Two Ways to Put AI in Your Books
AI on top of your books, or AI-native books? The two 2026 patterns compared with named vendors, real prices, and what each means for owners.
Read post

VisionAugust 11, 2026

## Adding an Industry to an ERP: a Recipe, Not a Kitchen
Why a new vertical in an AI-native ERP is a manifest rather than model training, what a generated tattoo-parlour module actually required, and where the analogy breaks.
Read post

EducationAugust 7, 2026

## ChatGPT for Accounting: Where the Ledger Stops It
ChatGPT is good at deciding what a transaction means and incapable of being your books. The difference is a system of record, invariants, ownership, and tests.
Read post

ComparisonJuly 28, 2026

## A2X vs Synder: Which Shopify Connector Fits in 2026
A2X vs Synder for Shopify sellers, compared honestly: posting methods, current pricing, reconciliation workflow, and the third path that skips the bill.
Read post

ComparisonJuly 13, 2026

## Free AI Accounting Tools 2026: What Free Buys You
Free AI accounting tools in 2026: Wave, Zoho Books, Puzzle, Odoo, GnuCash and ERPClaw compared, including what each free tier covers and where the paywalls sit.
Read post

ComparisonJune 30, 2026

## SAP Business One Alternatives 2026: Open-Source Picks
A vendor-neutral guide to real SAP Business One alternatives in 2026, from Acumatica and NetSuite to open-source ERP, plus a test for AI-native software.
Read post

ComparisonJune 16, 2026

## ERPNext Alternatives 2026: the AI-Native Take
A vendor-neutral guide to the real ERPNext alternatives in 2026, including the open-source options that are also AI-native, not just AI bolted on via plugins.
Read post

ComparisonJune 14, 2026

## Sage Intacct Alternatives 2026: the Open-Source Take
A vendor-neutral guide to the real Sage Intacct alternatives in 2026, plus a simple test for telling genuine AI-native finance software from AI bolted on.
Read post

ComparisonJune 10, 2026

## NetSuite Alternatives in 2026: The Honest, Open-Source Take
A vendor-neutral guide to the real NetSuite alternatives in 2026, plus a simple test for telling genuine AI-native ERP from AI bolted onto old software.
Read post

EducationJune 9, 2026

## Self-Hosted AI Accounting: Books on Hardware You Own
What self-hosted AI accounting actually means, why data sovereignty matters for your books, what it looks like day to day, and when self-hosting is the wrong call.
Read post

VisionMay 30, 2026

## The State of AI in Accounting: A 2026 Outlook
Where AI actually shipped value in accounting by 2026, what underperformed the hype, the automation tooling stack that earns its place, and what comes next.
Read post

ComparisonMay 30, 2026

## When Xero and QuickBooks Stop Fitting (and What Comes Next)
Xero and QuickBooks work well for most small businesses until they don't. Five growth triggers that signal you've outgrown them, and the honest alternatives in 2026.
Read post

VisionMay 30, 2026

## Where AI Belongs in Accounting (And Where It Doesn't)
Three places AI earns its keep in accounting and three places it doesn't. A practitioner framework with the decision tree, not the marketing pitch.
Read post

ComparisonMay 28, 2026

## Open Source AI Accounting: The 2026 Honest Roundup
2026 roundup of open source AI accounting tools. ERPClaw, ERPNext, Akaunting, Manager.io, Odoo. Which is AI-native, which is bolt-on, honest gaps.
Read post

ComparisonMay 23, 2026

## ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared
ERPNext core ships no native AI in 2026. The third-party Frappe app ecosystem filled the gap with NextAI, ChatNext, ChangAI, and an MCP bridge. Here is the map.
Read post

ComparisonMay 23, 2026

## Odoo Enterprise AI Pricing 2026: Plans, Features, and Cost Math
A dated Odoo Enterprise AI pricing snapshot with plan scope, feature access, and one, three, and five year cost math for small teams.
Read post

ComparisonMay 23, 2026

## SAP Joule and Business One: Embedded or Not?
SAP Joule is the headline AI inside S/4HANA Cloud, SuccessFactors, and Ariba. It is not inside SAP Business One. Here is what that means for an SMB founder in 2026.
Read post

ComparisonMay 8, 2026

## The 5 AI-Native ERPs That Actually Earn the Label
Most products tagged AI-native are AI-decorated. The five ERPs that actually rebuilt around AI agents, ranked by scope and structural moat.
Read post

ComparisonApril 26, 2026

## Odoo Alternative: When an AI-Native, Open Source ERP Fits
Evaluating Odoo alternatives? Compare tradeoffs in price, AI architecture, databases, implementation, manufacturing depth, and migration readiness.
Read post

EducationApril 25, 2026

## Double-Entry Bookkeeping for Developers, in Code
Double-entry bookkeeping explained the way developers think: invariants, immutable logs, and why debits equal credits, with code.
Read post

EducationApril 24, 2026

## FIFO vs Weighted Average Inventory: COGS and Profit
Compare FIFO and weighted average inventory costing with one rising cost example, a decision guide, and the effects on COGS, profit, tax timing, and ending inventory.
Read post

EngineeringApril 23, 2026

## Shopify OAuth on Cloudflare Workers: Our Pairing Bridge
How we built a Cloudflare Worker bridge between Shopify's embedded App Bridge OAuth and a self-hosted CLI ERP, with the gotchas we hit along the way.
Read post

VisionApril 21, 2026

## ERPClaw Q2 2026: What Shipped, What's Next
The first quarterly state of the platform. Two deep integrations live, 73 pages of marketing site, a rebuilt persona-driven demo, and what is next.
Read post

EducationApril 18, 2026

## Three Way Matching in Procure to Pay: A Plain English Guide
Three way matching catches paying for goods you never received and overpaying invoices. How the PO, receipt, and invoice match works, in plain English.
Read post

ComparisonApril 15, 2026

## The Sage Intacct Alternative That Costs $0
Sage Intacct pricing is quote-gated and climbs with modules. ERPClaw posts a real general ledger, covers the whole business, and costs $0 forever.
Read post

EducationApril 11, 2026

## Agency Accounting, Project P&L, and Time Billing
Why your friendly retainer client is secretly losing you $40 an hour, the four metrics every agency owner should know, and how to fix project P&L.
Read post

ComparisonApril 7, 2026

## The NetSuite Alternative for Teams Who Saw the Quote
NetSuite is the default mid-market ERP and the quote shows it. ERPClaw is the AI-native open-source alternative at roughly 100x less. An honest comparison.
Read post

EngineeringApril 3, 2026

## SQLite or PostgreSQL: Why ERPClaw Defaults to SQLite
ERPClaw is database agnostic via PyPika. SQLite is the default for self-hosted installs, PostgreSQL is fully supported for enterprise workloads. Here is why.
Read post

LaunchMarch 29, 2026

## ERPClaw's Shopify Integration: Free and Live Today
ERPClaw's Shopify integration is free, open source, and installable today: orders, refunds and payouts posted into real books.
Read post

ComparisonMarch 25, 2026

## Free Shopify Accounting Software in 2026: The Honest Guide
A real comparison of free Shopify accounting software in 2026. What is actually free, what is freemium with a paywall, and what works for a small store.
Read post

ComparisonMarch 21, 2026

## Webgility Alternative: Free Shopify ERP, QBO Included
Webgility runs $59 to $499 a month to sync Shopify into QuickBooks. ERPClaw replaces both for $0: the sync and the books, in one system.
Read post

EducationMarch 17, 2026

## Shopify Payout Reconciliation: Why It Never Matches
Your Shopify deposit doesn't match your sales total. Here's why Shopify payout reconciliation is a mess, and the simple pattern that fixes it.
Read post

EducationMarch 13, 2026

## Shopify Gift Card Accounting: the Deferred Revenue Trap
Gift card sales aren't revenue, they're a liability. Here's how Shopify gift card accounting actually works, why most stores get it wrong, and the fix.
Read post

ComparisonMarch 9, 2026

## Bookkeep Alternative: Daily Summaries Cost You Later
Bookkeep is a cheap Bookkeep alternative to A2X, but daily summaries cost you at audit time. ERPClaw posts every Shopify order to your books, free.
Read post

EducationMarch 5, 2026

## Stripe Connect Application Fees: Accounting and Journal Entries
Accounting for Stripe Connect application fees in direct charges, destination charges, and separate charges and transfers, with worked journal entries and refund entries.
Read post

ComparisonMarch 1, 2026

## Synder Alternative: One Tool Instead of Synder + QBO
Synder syncs your Shopify sales into books you still pay for. ERPClaw is the books: free, open source, with the Shopify integration built in.
Read post

EngineeringFebruary 25, 2026

## Building an AI-Native ERP With Claude Code
How a spec-first methodology, constitutional financial laws, and Claude Code shipped a full back-office open source ERP a large team would take years to build.
Read post

EducationFebruary 21, 2026

## MRR vs ARR in Stripe: Definitions, Formula, and Example
Calculate MRR and ARR from Stripe subscriptions, with formula boundaries, annual plan normalization, status rules, a worked example, and the difference from recognized revenue.
Read post

ComparisonFebruary 17, 2026

## A2X Alternative: Free, Open Source, Built for Shopify
Looking for an A2X alternative? ERPClaw uses the same clearing account method, books every order separately, includes per-warehouse stock costs, and costs $0.
Read post

EducationFebruary 13, 2026

## The Clearing Account Pattern, in Plain English
The clearing account pattern is the trick that makes Shopify and Stripe deposits actually balance to your sales. Here is how it works, in plain English.
Read post

LaunchFebruary 9, 2026

## ERPClaw + Stripe: Deep Coverage, ASC 606, $0
ERPClaw's Stripe integration is in production: payout reconciliation in three layers, ASC 606 revenue recognition, and Connect fees. Free and open source.
Read post

ComparisonFebruary 5, 2026

## ERPNext vs ERPClaw: an Honest Open-Source Comparison
ERPNext is a proven open-source ERP with a 15-year track record. ERPClaw is AI-native, modular, open source. Honest tradeoffs from a founder who has shipped both styles.
Read post

EducationFebruary 1, 2026

## Shopify Cost of Goods Sold: Your Real Margin Number
Most Shopify stores miscalculate cost of goods sold by 10-20%. Here is what actually counts as COGS, what Shopify misses, and how to fix it for free.
Read post

ComparisonJanuary 27, 2026

## Rillet Alternative: Free AI Accounting for Stripe
Rillet quotes run $2,000+/month (founder-reported). ERPClaw does AI-native Stripe accounting inside a full ERP for $0, open source.
Read post

VisionJanuary 23, 2026

## Open Source vs SaaS: The ERP Question Most Buyers Get Wrong
SaaS ERP vendors charge per user forever and own your data. Open source gives you the code and runs on your machine. The honest tradeoff, either way.
Read post

ComparisonJanuary 19, 2026

## The FreshBooks Alternative for When Freelance Becomes Agency
FreshBooks is great until you hire staff, take on inventory, or hit the 51-client tier. Here is the open-source FreshBooks alternative that scales with you.
Read post

EducationJanuary 15, 2026

## Stripe ASC 606 Revenue Recognition: A Founder's Guide
ASC 606 for SaaS founders on Stripe: the five-step model, journal entries, mid-cycle changes, refunds, trials, and an audit-prep checklist.
Read post

ComparisonJanuary 11, 2026

## Open Source Xero Alternative: Own Your Books, Pay $0
Xero is great cloud accounting. Add Payroll, Projects, Expenses, and an inventory plugin and you are at $300+/mo. The open-source ERP that bundles it free.
Read post

VisionJanuary 7, 2026

## AI-Decorated vs AI-Native: Why Most AI Features Lose
AI-decorated tools bolt a chatbot onto 2015 software and charge a new fee. AI-native software rebuilds the architecture. One of these wins. Here is why.
Read post

ComparisonJanuary 3, 2026

## The QuickBooks Alternative That's Actually a Full ERP
A QuickBooks alternative that is a full ERP: accounting, inventory, and payroll in one free, open-source, self-hosted system.
Read post

VisionDecember 31, 2025

## From SAP Consultant to Open Source ERP Builder
18 years across enterprise software, six of them rolling out SAP at Accenture. Here is why I left, what I learned, and why ERPClaw exists now.
Read post

## Stay updated
Get notified when we publish new posts. No spam, unsubscribe anytime.

---

## The 5 AI-Native ERPs That Actually Earn the Label
URL: https://www.erpclaw.ai/blog/5-ai-native-erps-that-earn-the-label/

# The 5 AI-Native ERPs That Actually Earn the Label
Most products tagged AI-native are AI-decorated. The five ERPs that actually rebuilt around AI agents, ranked by scope and structural moat.

“AI-native ERP” is the most-claimed term in enterprise software in 2026. SAP says Joule. Oracle says AI Agents. Microsoft says Copilot. NetSuite, Sage Intacct, Xero, QuickBooks, Odoo, ERPNext: all claim AI. Almost none of them rebuilt the architecture. They added a chat sidebar to a forms-and-workflows product designed in 1998 and called it AI-native.

A small group of products genuinely earned the label. They were designed from commit one for an AI agent to be the primary user, not a click-through human. They enforce data integrity at write time, not at audit time. Their action layer is the API, the UI, and the integration surface all at once.

This piece ranks the 5 we think actually earn the label. Order is by scope and structural moat depth, not by buzz or funding. We will explain the test we used, name names, and call out the honest gaps in each.

If you only have 30 seconds: ERPClaw is the broadest scope, the only open-source option, and the only one that’s $0 forever. Rillet is the strongest finance-close-only AI-native SaaS. Doss is the agent-orchestrated workflow option. Campfire is the lightweight startup choice. DualEntry is the migration-friendly finance pick.

Read the full pillar at AI-native ERP: the 5-trait test for the architectural framework. This post is the head-to-head ranking that uses that framework.

## What “AI-native” actually means

The 5-trait test (full version on the pillar):

- User experience. Conversational and proactive, not menus and forms.

- Workflow. Agentic reasoning and action, not rule-based approvals plus batch jobs.

- Data. Semantic and explainable, not just rows in tables and BI dashboards.

- Automation. AI is embedded in the modules, not external scripts and RPA bots.

- Governance. Role security plus AI controls, traceability, model governance, human approval gates.

A vendor is AI-native if every layer is agent-first by design. AI-decorated if the AI is a paid module bolted onto a product that shipped before ChatGPT.

The 30-second buyer test: read the vendor’s product page, then ask their sales team “can your AI post a journal entry on its own, with no human in the loop, on a workflow you didn’t pre-build for it?” If the answer is no or “we have approval gates”, they are AI-decorated. If the answer is yes plus governance, they are AI-native.

By that test, NetSuite Joule fails. Oracle AI Agents fail. Microsoft Dynamics Copilot fails. Sage Copilot fails. QuickBooks Intuit Assist fails. Xero Just Ask fails. Odoo’s plug-in stack (changAI, NextAI, Ollama, Composio) fails. ERPNext fails by the same measure. So does pasting your books into a general chatbot: ChatGPT for accounting covers exactly where that approach stops working.

Five products pass. Here they are.

## 1. ERPClaw, broadest AI-native ERP

ERPClaw is the AI-native ERP we built. Open-source under GPL v3, self-hosted by default, free forever. It covers the full operations surface across 14 industry verticals, with a deep action and schema surface backed by a constitutional test suite that enforces 18 financial laws at write time.

Why it earns the AI-native label:

- Action layer is the API. Every business operation, from `add-customer` to `submit-sales-invoice` to `reconcile-stripe-payout`, is a discrete action callable by an AI agent. There is no UI dependency. The web dashboard, when it ships, is a thin wrapper over the same actions a chat agent already uses.

- Constitutional invariants enforced at write time. Double-entry balance, voucher balance, immutability of submitted GL entries, and the rest of the financial rule set are checked on every transaction. Mutability bugs that QuickBooks shipped for 20 years are structurally impossible.

- Spec-first regen. Every module has a SKILL.md spec. Any LLM can regenerate the code from the spec. Customization is not a SuiteScript engagement. It is a prompt.

- Database-agnostic. SQLite by default with WAL mode and FK enforcement, PostgreSQL fully supported via PyPika. Same code on either. No data lock-in.

Honest gaps:

- The web dashboard (webclaw) is developer-only today. Most users interact via Claude or another AI agent in their terminal.

- Support is community-tier (GitHub issues + Discord) not 24/7 enterprise.

- Implementation services are available from AvanSaber Inc. but the product is built so most teams self-onboard in five minutes.

Pricing: $0 forever. Self-hosted on your machine. No SaaS subscription, no per-seat fee, no metered usage.

Where to read more: /ai-native-erp/, /features/, /docs/core/install/, or /compare/netsuite/.

## 2. Rillet, AI-native finance close

Rillet is the strongest commercial AI-native finance SaaS in 2026. Founded by Nicolas Kopp and team (per TechCrunch, May 2025). Closed-source, vendor-cloud-only, seat-based pricing. Founder-reported quotes land in the $2,000 to $10,000 per month range.

Why it earns the AI-native label:

- Built around the close cycle. Their chat agent reconciles, posts journal entries, generates board-deck-ready financials. AI is the primary surface, not a sidebar.

- Variance analysis and anomaly detection are first-class, not a separate report. The agent surfaces them in the close workflow.

- Modern stack. Designed for an AI agent first.

Honest gaps versus ERPClaw:

- Finance only. No inventory, manufacturing, projects, HR, payroll, regional tax modules. If your company is past pure finance, you also need other systems.

- Closed source and SaaS-only. Your books live on Rillet’s servers. Migration off is the SaaS migration story you already know.

- Seat-based pricing. A growing finance team scales the bill linearly.

- Limited integrations breadth versus a full ERP. Happy users on G2 consistently wish for more module reach.

When to pick Rillet over ERPClaw: you are a large SaaS company, you don’t need a full ERP, you want managed finance close software, and your finance team prefers a SaaS dashboard to a chat agent.

We have a full head-to-head at /compare/rillet/.

## 3. Doss, agent-orchestrated ERP

Doss is the AI-native option built around agent-orchestrated workflow configuration. Closed-source, SaaS, narrower module set, pricing-opaque (contact sales).

Why it earns the AI-native label:

- Agent-driven workflow setup. Onboarding is a chat with the agent that configures the system, not a 90-day implementation engagement.

- Strong on rapid deployment. A new tenant goes live faster than an Odoo install or a NetSuite rollout.

Honest gaps versus ERPClaw:

- Closed source. No code review, no fork, no self-host.

- SaaS-only. Vendor cloud lock-in.

- Narrower scope. Not a full multi-vertical ERP.

- Pricing opaque. Contact sales is a tax on the buyer, not a feature.

When to pick Doss over ERPClaw: you want a SaaS-managed AI-native ERP for a single-vertical use case where the Doss feature set fits, and you don’t need open source.

## 4. Campfire AI, lightweight AI for startups

Campfire is the AI-native pick if you are a 2 to 20 person startup that wants a chat-driven finance and operations stack without standing up infrastructure.

Why it earns the AI-native label:

- Chat-first onboarding. No 12-screen forms wizard. Talk to the agent.

- Designed for the early-stage founder. Light scope, high speed.

Honest gaps versus ERPClaw:

- Scope ceiling. Will not carry you past Series A operations the way a full-scope ERP does.

- Closed source and SaaS. Same lock-in story as Rillet and Doss.

- No on-prem option. If your investors or auditors require self-hosted, Campfire is out.

When to pick Campfire over ERPClaw: you are a non-technical founder, you want a managed-SaaS AI experience right now, and you don’t have a developer or AI agent set up to drive ERPClaw.

## 5. DualEntry, AI-native finance migration

DualEntry rounds out the list as the AI-native finance pick optimized for migration off legacy accounting tools. Their wedge is “switch off QuickBooks or Xero in a day, with the AI handling the transcription work.”

Why it earns the AI-native label:

- AI-driven migration. The agent reads your existing QBO or Xero export and rebuilds the books on DualEntry without a human re-keying anything.

- Finance-first action layer. Like Rillet, the AI is the primary user.

Honest gaps versus ERPClaw:

- Finance only. No inventory or manufacturing or HR.

- Closed source, SaaS, seat-based. Standard SaaS economics.

- Smaller team and shorter track record than the rest of the list. The migration wedge is real but the vendor is earlier.

When to pick DualEntry over ERPClaw: you are migrating off a legacy finance SaaS this quarter, you want a managed migration done by AI, and you accept a finance-only scope.

## How to evaluate AI-native vs AI-decorated

Five questions to ask any vendor before you sign:

- Can your AI post a journal entry without a human in the loop? Yes plus governance is AI-native. No or “we have approval gates” is AI-decorated.

- Is the AI the API, or is the UI the API? If the AI is calling the same endpoints a click would call, the AI is bolted on.

- Where does the AI run when there is no chat session open? AI-native systems have agents posting transactions on a schedule. AI-decorated systems sit idle until a human opens chat.

- What invariants are enforced at write time, not at audit time? AI-native architectures enforce double-entry balance, immutability, and constitutional rules on every commit. AI-decorated tools rely on after-the-fact reconciliation.

- Can a customer regenerate the source code from a spec? AI-native systems treat code as derivable from a SKILL.md. AI-decorated systems are a 20-year-old codebase wrapped in copilot prose.

If a vendor answers no on three or more, they are AI-decorated. They might still be a good product. They are not AI-native.

The full framework is on /ai-native-erp/. For the prior question, which parts of the accounting workflow AI should own in the first place, where AI belongs in accounting draws the line.

## A note on Versori’s list

Versori published a list of 5 AI-native ERPs. It is a useful starting point. We disagree on one entry: Versori included Opkey, which is an ERP testing automation platform, not an ERP. We swapped Opkey for DualEntry, which is genuinely an AI-native finance product.

Editorial honesty matters here because most listicles in this category recycle each other’s vendor picks without verification. The 5 names on this list earn the label by the 5-trait test above. If you can show us a sixth that passes the same test, we will add it.

## Frequently asked questions

Is ERPClaw really an ERP, or just an accounting tool?

ERPClaw is a full ERP. It covers accounting, inventory, manufacturing, HR, payroll, CRM, billing, AI engine, projects, fleet, treasury, and 14 industry verticals. The accounting and finance modules are rock-solid (immutable GL, ASC 606 revenue recognition, 7 currencies, 12-step posting validation), but the breadth is the structural moat.

What’s the difference between AI-native and AI-decorated?

AI-decorated software is a forms-and-workflows product from 1998 to 2015 with a chat sidebar bolted on after ChatGPT. The AI is a feature. AI-native software is built around the AI agent from commit one. The AI is the primary user. The architecture, data model, governance, and update cadence all assume the agent is driving meaningful writes. We cover this in detail on AI-native ERP.

Can ERPClaw run on PostgreSQL?

Yes. ERPClaw uses PyPika, a SQL query-builder abstraction. Same code targets SQLite or PostgreSQL with no per-database forks. SQLite is the default because most teams under 100 users do not need a separate database server. PostgreSQL is fully supported and recommended for teams that already run Postgres or expect very large transaction volumes.

How is ERPClaw free?

ERPClaw is open source under GPL v3. The software is $0 forever because there is no SaaS company between you and the code. AvanSaber Inc., the company behind ERPClaw, makes money from optional implementation services, hosted ClawHub for teams that prefer not to self-host, and from enterprise support contracts. The product is free; the services around it are optional.

Is Rillet, Doss, Campfire, or DualEntry a better fit for me than ERPClaw?

Sometimes yes. Pick Rillet if you are a large revenue SaaS company that wants managed finance close software and only needs finance scope. Pick Doss if a managed-SaaS AI ERP fits your single-vertical use case. Pick Campfire if you are an early-stage founder who wants chat-driven finance without setting up infrastructure. Pick DualEntry if you are migrating off QBO or Xero this quarter and want the AI to handle the transcription. Pick ERPClaw if you want full ERP scope, open source, self-hosted, and $0 forever.

When would I NOT pick ERPClaw?

If you need a polished web dashboard today (webclaw is dev-only), if your finance team will not adopt a chat-first workflow, if you need 24/7 vendor-managed enterprise support, or if you are a Fortune 500 with multi-entity consolidation requirements that exceed our current 7-currency single-invoice model. Those are real constraints. The honest answer is “ERPClaw is built for solo founders to mid-market, not Fortune 500.”

## Where to go next

Read the framework: /ai-native-erp/ for the 5-trait test in full.

See the matrix: /erp-comparison/ for a 13-vendor side-by-side that includes the AI-decorated incumbents this list excluded.

Try the demo: /demo/ gives you a live chat with ERPClaw without setup.

Install: /docs/core/install/ gets you running in five minutes.

Compare directly: /compare/rillet/ is the head-to-head with the strongest commercial AI-native peer on this list.

---

## A2X Alternative: Free, Open Source, Built for Shopify
URL: https://www.erpclaw.ai/blog/a2x-alternative-erpclaw-comparison/

# A2X Alternative: Free, Open Source, Built for Shopify
Looking for an A2X alternative? ERPClaw uses the same clearing account method, books every order separately, includes per-warehouse stock costs, and costs $0.

If you sell on Shopify and you have ever Googled “A2X alternative,” you probably ended up on Synder, Bookkeep, or Webgility. They are all paid apps. They all start somewhere between $25 and $50 a month, then climb fast as you add stores, more orders, or the stock-cost tracking you actually need.

Almost nobody points you at the free, open-source option. So this post does.

ERPClaw is a free, AI-native, open-source ERP with a Shopify integration that is shipped and installable today. It uses the same clearing-account method A2X invented. It books every order separately so you can trace every dollar. It tracks per-warehouse stock costs in the free version. And it costs zero dollars a month, forever, because it runs on your own computer or server.

A2X is excellent at what it does, but it is AI-decorated software (the assistant is a chat sidebar bolted on the side). ERPClaw is the only AI-native option in this category, which means the assistant is the primary interface, the spec drives the code, and the architecture was built around AI from day one. That is a structural difference no Shopify accounting connector built before 2023 can retrofit.

This is the long version of the comparison page at /compare/a2x. I am writing it because A2X is a good product, the team built something genuinely useful, and any fair comparison has to start there before it gets to where they fall short.

## What A2X gets right (and why most stores already use it)

Before A2X, Shopify bookkeeping was a mess. Stores would either:

- Type every order one by one into QuickBooks and watch their accountant cry, or

- Dump a daily total into a single line and lose all the per-order detail, or

- Pay an offshore bookkeeper to match payouts in a spreadsheet every Friday.

A2X invented the now-standard fix. Every Shopify sale flows through a holding bucket called a clearing account. When Shopify pays out to your bank, the clearing account empties. The math always nets to zero, the bank deposit always matches, and the auditor always smiles. This is the right setup. Every other tool in the category copies it now, including ERPClaw.

A few other things A2X genuinely got right:

- Trust. They have been doing this since 2014. The product is battle tested across Shopify, Amazon, Walmart, eBay, and Etsy.

- QuickBooks Online and Xero plugs. Deep, mature, well documented. If your accountant lives inside QuickBooks and refuses to leave, A2X plugs in cleanly.

- Multi-channel. One subscription handles Shopify and Amazon and a few others without you reinventing the chart of accounts each time.

- Onboarding polish. The setup wizard is one of the cleanest in the category.

- Documentation. The A2X help center is a teaching resource, not just a help center. I have learned things from it.

I am happy to credit them this much because the rest of the post is going to be less polite. I want it on the record that the team did real work.

## Where A2X falls short

A2X is not free. That is the headline. The deeper problems are structural.

### 1. The price climbs fast and locks features behind tiers

A2X starts at around $29 a month on the A2X pricing page (as of August 2026) for a small Shopify store. That sounds reasonable until you read the small print. The $29 tier caps you at a low order count. Once you grow, you bump to $45, then $79, then through the Premium tiers at $115, $159, and $229 a month, with higher-volume tiers above that. Each tier opens up features that the next tier still keeps locked.

The features behind the paywalls are the ones store owners actually want:

- Cost-of-goods tracking is locked above the entry tier.

- Multiple currencies is locked.

- Multiple stores is locked.

- Tracking stock by location is locked.

If you are a real Shopify store doing real volume with real product variants, you are paying $79 to $229 every month, forever, just to see the profit on each shirt you sell. Across five years that is $4,740 to $13,740 in fees for software that types numbers into your books.

### 2. It is a connector, not your books

A2X is a sync layer. It moves data from Shopify into QuickBooks Online or Xero. That means you still need:

- QuickBooks Online (around $30 to $200 a month) or Xero (around $15 to $80 a month) as the actual books.

- A separate stock app like Cin7, DEAR, or Katana once you outgrow QuickBooks’s built-in stock features.

- A separate customer-tracking tool.

- A separate payroll tool.

- A separate bill-pay tool.

Your $29 A2X bill is the entry fee. The full stack to run an ecommerce business properly with A2X in it is closer to $300 a month before you have hired anyone or shipped a single product.

### 3. No real stock module

A2X exports stock data so you can hand it to QuickBooks. It is not a stock system. There is no idea of warehouses, bin locations, transfers, cycle counts, or a stock log that matches your books to the cent. If you ship out of more than one location, A2X cannot tell you what is sitting at each one or what each shirt costs you on average per warehouse.

### 4. Daily summaries by default

A2X defaults to one big lump per day. One bookkeeping line covers a whole day or a whole payout. That is fine until a customer disputes a single charge and you have to find which order, which line item, which tax row. With daily summaries you go back to Shopify, find the order, do the math by hand, and trust your spreadsheet. Booking each order separately does not have this problem because every order has its own entry with the full Shopify details attached.

### 5. Closed source

You cannot read A2X’s revenue-recognition code. You cannot inspect how they handle a partial refund on a multi-line order with a gift card and three discount codes. You take the entry on faith. For a tool whose whole job is your books of record, that is a strange amount of trust to extend to a black box.

### 6. Stuck in their cloud

Your historical mappings, your custom rules, your category logic all live on A2X servers. If they raise prices, get acquired, or kill a feature, your data comes back to you as a CSV and you start over.

## What ERPClaw does instead

ERPClaw uses the same clearing-account method A2X pioneered. From there it goes a different direction in four ways that matter.

Every order, every time. Every Shopify order becomes a real sales invoice in ERPClaw with line items, taxes, shipping, discounts, and the full Shopify order details attached. Refunds become credit notes pointed at the original invoice. Disputes get their own entry. When the auditor asks “what is this $847.32 line in your revenue?” you click through to a single order.

Per-warehouse stock costs in the free version. ERPClaw ships with a real stock module. Stock costs work the way most accountants want (oldest stock sold first, called FIFO). Multi-warehouse stock log. Bin locations. Cycle counts. Transfers. The stock log ties to your books, every entry, every minute. Not a paid upgrade. Just included.

It is your books, not a connector to someone else’s. ERPClaw covers the full back office. Sales, purchasing, stock, manufacturing, HR, payroll, customer tracking, projects, fixed assets, fund accounting, the works. The Shopify integration is one of those modules. You do not need QuickBooks Online underneath. ERPClaw is your books.

Free, open source, you host it. open source. Source on github.com/avansaber/erpclaw. Runs on your laptop, your office Mac mini, your $5 VPS, or a beefy server. Your data lives where you put it. We never see your orders, customers, or money. Our small Cloudflare connector at `shopify.erpclaw.ai` is just a pairing helper and a status mirror, all documented in the Shopify architecture doc.

AI-native, not AI-decorated. A2X has a polished web UI built in 2014 and a chat helper bolted on later. ERPClaw is operated through your AI assistant in plain English. “Pair my Shopify store, pull the last 12 months of history, then show me April’s gross margin by SKU” is one sentence, not five clicks across three screens. The AI is load-bearing, not a sparkle icon. The full argument is in AI-decorated vs AI-native software.

## Side by side

| Feature
| A2X
| ERPClaw

| Price
| $29 to $229 a month
| $0 forever (open source license)

| Payout matching
| Yes (clearing account)
| Yes (clearing account, 3-layer)

| Books each order separately
| Daily summary by default
| Every order, every time

| Cost-of-goods tracking
| Premium tier ($49+)
| Built in, multi-warehouse

| Gift card revenue handling
| Premium tier
| Built in

| Stock module
| No (export only)
| Yes (full system, multi-warehouse)

| HR / Payroll / Customer tracking
| No
| Yes (in the same app)

| Run on your own server
| No
| Yes

| Source code access
| Closed
| Open on GitHub

| Architecture
| AI-decorated (chat sidebar bolted on)
| AI-native (assistant is the primary interface)

| Database
| Their cloud (proprietary)
| SQLite or PostgreSQL on your machine (PyPika abstracts the layer)

| Stripe integration
| No
| Yes (separate module)

| Custom posting rules
| No
| Yes

| Multiple currencies
| Yes
| 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED), payment-side

| Needs QuickBooks or Xero
| Yes
| No (ERPClaw is your books)

On currencies the two differ in shape rather than winner. ERPClaw supports 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED) on the payment side: an invoice settles in the currency it was raised in, and no conversion ever happens inside a transaction. A2X leans on the FX conversion of the books it posts into. A UK or EU store runs fine on either; they just handle FX differently.

## How to switch (the short version)

Switching from A2X to ERPClaw is a one-evening project for most stores. The full step-by-step guide is being written and will live at `/docs/migration/from-a2x` when ready. Here is the shape of it.

Step 1. Pick a switch-over date. Usually the first day of a fresh accounting period. End of month is cleanest.

Step 2. Install ERPClaw on the box you want it to run on. Five minutes start to finish. The full setup is at /docs/shopify/install-walkthrough.

Step 3. Pair the ERPClaw Shopify integration. The integration is shipped and installable today; the feature tour shows what it does and the docs cover setup. The embedded admin shows a pairing code. Run `erpclaw shopify-connect --pairing-code ABC-DEF` on your ERPClaw box. Done. There is also a self-hosted custom-app token flow if you prefer that route.

Step 4. Pull in your history. ERPClaw will pull every order, refund, payout, and dispute from Shopify going back as far as you tell it. Most stores pull the last 12 months. The history sync runs once and books every transaction with its original date.

Step 5. Match the books on switch-over day. Compare ERPClaw’s trial balance on the cutover date against the trial balance A2X handed to QuickBooks. They should match within rounding. If they do not, the three-layer reconciliation report tells you exactly which payout, which order, which line is off.

Step 6. Cancel A2X. Save a copy of your A2X mappings as a record. You will not need it again, but auditors love a paper trail.

Step 7. Decide the QuickBooks question. Most stores drop QuickBooks entirely once ERPClaw is the books. Some keep it running in parallel for a quarter to make their accountant comfortable. ERPClaw can push a one-way summary export to QuickBooks for that hybrid period. You sunset QuickBooks whenever you are ready.

That is the playbook. A long Saturday afternoon, max.

## Free is not just the price tag

Worth saying plainly because the headline does most of the work and people miss the rest. ERPClaw is free, not because there is a paid tier we are pushing you toward. There is no paid tier. There is no enterprise version. There is no “premium support” SKU. The full product is free. Forever.

The business model is documented at /pricing. Short version: we make money on managed cloud hosting (for stores who do not want to host it themselves), industry-specific consulting, and eventually a marketplace for third-party add-ons. None of that locks the core product. The core product is the entire app, every module, including the Shopify integration, including stock costs, including the books themselves.

If that sounds too good to be true, read the source. That is the point of an open license.

## FAQ

Is ERPClaw really free, or is there a catch?

Really free. Open source. No paid tier. No upsell to a “pro” version. The catch, if you want to call it one, is that you host it. You run it on your own machine. If you would rather pay someone to host it for you, ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/. The self-hosted version is and will always be free.

Does ERPClaw work with QuickBooks Online?

ERPClaw replaces QuickBooks Online. It is your books, not a connector to someone else’s. If you want to keep QuickBooks running in parallel during a switch-over period, ERPClaw can push a one-way summary export to it. Most stores drop QuickBooks once they see ERPClaw’s reports.

What about Xero?

Same answer. Xero is a set of books. ERPClaw is a set of books. You do not need both, and if you want a hybrid setup during a switch, the per-business Xero sync keeps both current while you decide.

Does ERPClaw handle multiple currencies the way A2X does?

Differently. ERPClaw supports 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED), payment-side: each invoice settles in the currency it was raised in, with no conversion inside a transaction. If your store specifically needs AUD books, A2X paired with a local ledger covers that today; for USD, EUR, GBP, CAD, INR, or SGD, ERPClaw keeps the books itself.

Will booking every order slow my store down?

No. Each Shopify webhook gets booked within a few seconds. A high-volume store doing 10,000 orders a month will see 10,000 sales invoices in ERPClaw at month end. SQLite handles that comfortably. Reports run in milliseconds. The only slow part is the first history pull, which takes a few minutes per 1,000 orders.

Can I read the code that creates each entry?

Yes. Every posting rule is in the source repo. The 12-step validation, the clearing-account method, the cost-of-goods math, the gift card revenue logic, the dispute reserve handling. Read it, fork it, change it, audit it. That is the whole point of open source for accounting software.

## Closing

A2X is a good product. It is also a $29 to $229 a month bill that locks cost-of-goods, stock, and multi-store behind tiers, requires QuickBooks Online underneath, posts daily summaries instead of per-order entries, and keeps its source code closed.

ERPClaw is the free, open-source alternative that uses the same clearing-account method, books every order, ships with multi-warehouse stock costs in the free version, and is your books itself. No QuickBooks underneath. No tiers. No vendor lock-in.

If that sounds like what you have been looking for, here is where to start:

- Get the Shopify integration: Shopify integration (installable today; App Store listing pending)

- Read the install walkthrough: /docs/shopify/install-walkthrough

- See the architecture: /docs/shopify/architecture

- See the comparison page: /compare/a2x

- Read the source: github.com/avansaber/erpclaw

- Pricing: /pricing (spoiler: it is free)

If you have questions, email `support@avansaber.com`. If you find a bug, file it on GitHub. If you switch from A2X and want to write up your migration story, I will publish it on this blog with full credit.

You do not have to keep paying a monthly bill to keep your Shopify books clean. Try the free option for one weekend. If it does not save you money, walk away.

---

## A2X vs Synder: Which Shopify Connector Fits in 2026
URL: https://www.erpclaw.ai/blog/a2x-vs-synder-shopify-accounting/

# A2X vs Synder: Which Shopify Connector Fits in 2026
A2X vs Synder for Shopify sellers, compared honestly: posting methods, current pricing, reconciliation workflow, and the third path that skips the bill.

Short answer. A2X and Synder do the same job in two different shapes. A2X posts one summary journal entry per Shopify payout that nets exactly to your bank deposit: pick it if you keep accrual books and want reconciliation to be a one-line match. Synder posts every order, refund, and fee as its own transaction (with a daily-summary mode as fallback) and covers 30+ sales channels in one subscription: pick it if you sell across several rails or need customer-level detail inside the ledger. Both are connectors, so the true monthly cost includes QuickBooks Online or Xero underneath. If the two-bill structure is the part you dislike, there is a third path: run the accounting where the orders land. We build ERPClaw, which takes that path, so weigh our framing accordingly.

We already keep long comparisons of each tool against our own (the A2X one, the Synder one), but the question sellers actually type is the direct one: A2X vs Synder, which do I buy? Fair question, and it deserves a straight answer rather than a detour. Prices below were checked against the vendors’ pages in July 2026, and the disclosure stands throughout: we make a competing product, and both of these tools are competent at what they do.

## The same job, two shapes

Both A2X and Synder read your Shopify activity and write it into accounting software you already run. Neither one is your books. QuickBooks Online or Xero is, and that single fact drives every difference below: because each tool posts into someone else’s general ledger, each had to decide how much detail that ledger can absorb without choking.

A2X chose summaries. Synder chose transactions. Almost everything else follows from that choice.

## How A2X works: one entry per payout

A2X groups everything inside a Shopify payout (orders, refunds, fees, adjustments, gift card movements) into a single summary journal entry that nets to the deposit that lands in your bank. A2X’s own description is that it “creates a summary entry” covering the exact payment period so the books match the bank, and it reports serving over 10,000 ecommerce businesses, accountants, and bookkeepers. Destinations are QuickBooks Online, Xero, NetSuite, and Sage. Reconciliation becomes a one-line match, which is why accountants who live in QBO tend to recommend it: the month closes on roughly one entry per payout instead of thousands of rows.

Shopify pricing as of July 2026, per the A2X pricing page: Mini at $29 a month covers 200 orders on one store, Basic at $45 covers 500, Professional at $79 covers 2,000, and the Premium tiers run from $115 (5,000 orders, up to 5 stores) to $1,039 (100,000 orders, up to 10 stores). Two gates worth knowing before you pick a tier: cost of goods sold and the Shopify gift card report start at Basic, not Mini. There is no permanent free tier; the trial lets you post three settlements with full features.

The honest weaknesses. Summary-first means per-order detail lives in A2X and Shopify, not in your ledger, so when a dispute needs order-level tracing you leave the books to find it. Multi-channel selling stacks costs, since each channel typically needs its own A2X subscription. And it is a connector, so the ledger bill underneath is separate and mandatory.

## How Synder works: every transaction, everywhere

Synder syncs each sale, refund, and fee into QuickBooks or Xero as its own record, or falls back to a daily summary mode if the volume gets heavy. Its distinctive strength is breadth: one subscription covers 30+ platforms, including Shopify, Amazon, eBay, Etsy, Walmart, Stripe, PayPal, Square, and WooCommerce, which is why multi-channel sellers gravitate to it. Smart Rules add if-this-then-that automation on top of the sync, with the allowance growing by tier.

Pricing as of July 2026, per Synder’s pricing page with the tier detail tracked by CostBench: Basic at $65 a month covers 500 synced transactions with 2 integrations and 1 user, Essential at $115 covers 3,000 transactions with unlimited integrations, Pro at $275 covers high-volume stores with 2 users, and larger tiers go up from there, with a custom Premium at the top. There is a 15-day trial and no permanent free tier.

The honest weaknesses. Synder meters synced transactions rather than orders, and refund and fee events ride along with sales, so the bill climbs faster than an order count suggests. Per-transaction posting floods QuickBooks at volume (a store doing 1,000 orders a month is pushing a few thousand ledger lines a month into QBO), and the common escape hatch, flipping to daily summaries, gives up the per-order granularity that justified choosing Synder in the first place. Reconciliation also takes an extra step: instead of one entry that nets to the deposit, you match deposits through a clearing account.

## Head to head

|
| A2X
| Synder

| Posting method
| One summary entry per payout
| Per transaction, or daily summary

| Bank reconciliation
| Entry nets to the deposit, one-line match
| Clearing-account step

| Books required underneath
| QuickBooks Online, Xero, NetSuite, Sage
| QuickBooks or Xero

| Channel coverage
| Per-channel subscriptions (Shopify, Amazon, eBay, Etsy, Walmart, PayPal)
| 30+ platforms in one subscription

| Metering unit
| Orders per month
| Synced transactions per month

| Entry price (July 2026)
| $29/mo, 200 orders
| $65/mo, 500 transactions

| Where the gates sit
| COGS and gift card report from the $45 Basic tier
| Smart Rules and users grow by tier

| Free tier
| No (trial posts three settlements)
| No (15-day trial)

| Best fit
| Accrual books, payout-match reconciliation
| Multi-channel sellers, customer-level detail

## The bill nobody quotes: the books underneath

Neither tool works alone, so the comparison that matters is stack against stack. QuickBooks Online runs from $38 a month for Simple Start to $275 for Advanced, with the popular Plus plan at $115 (as of July 2026). A mid-size store on A2X Professional plus QBO Plus is paying $79 plus $115, about $194 a month, before any other app. The same store on Synder Essential plus QBO Plus is at $115 plus $115, about $230 a month. Over a year the connector-plus-ledger stack lands somewhere between $2,300 and $2,800 for a store in that band, and it scales up with volume from there.

Nikhil Jathar, who co-founded ERPClaw at AvanSaber, argues the interesting line item is not either price:

“A2X and Synder both answer the same question well: how do I move commerce data into books that live somewhere else. The question worth asking in 2026 is why the books live somewhere else.”

## The verdict: which fits whom

- Pick A2X if you keep accrual books, sell mostly through Shopify (one or two channels), and want payouts to reconcile against the bank in one click. This is also the safe pick when your accountant lives in QBO or Xero and wants a clean month-end close; practitioner writeups like Ottit’s comparison land roughly 80/20 toward A2X for accrual DTC brands.

- Pick Synder if you sell across many rails and want one subscription instead of one A2X per channel, if you run cash-basis books, or if you genuinely need per-customer, per-order detail inside the ledger and your volume is low enough for QBO to absorb it.

- Either way, you are buying a connector plus a ledger, and both bills recur forever. That is a reasonable trade if the stack fits how your accountant works. Both products are mature; you will not pick a broken tool here.

## The third path: accounting where the orders land

The alternative neither vendor mentions is removing the sync entirely. ERPClaw is a free, open-source, AI-native ERP whose Shopify integration and Stripe integration write orders, refunds, fees, payouts, and gift card movements straight into its own general ledger. It uses the same clearing-account pattern A2X standardized, but because the ledger is local there is no third-party GL to flood: you keep per-order detail and the payout still nets to the deposit. Bank statement import (OFX, CAMT.053, MT940, BAI2) with an auto-match engine covers the deposit side, and a Plaid connector links bank accounts. The price is $0, self-hosted, GPL v3, with no order or transaction meter.

The honest catches, same as always: you install and host it yourself, the primary interface is an AI agent rather than a web dashboard your accountant already knows, and if your accountant insists on QBO, a connector really is the path of least resistance. If Bookkeep is also on your shortlist, we covered that corner in the Bookkeep alternative guide, and the wider option map lives in the free Shopify accounting software guide.

“A connector is rent on the gap between two systems. You can pay that rent forever, or you can close the gap.”

That is Jathar’s framing, and it is the fair way to size this decision: A2X and Synder are both good rent. ERPClaw is the option for sellers who would rather own the building.

## FAQ

Is A2X or Synder cheaper for a Shopify store?

At the entry level A2X, at $29 a month for 200 orders against Synder’s $65 for 500 synced transactions (both as of July 2026). But run the math on your own volume: A2X meters orders, Synder meters synced transactions (refunds and fees count), and both sit on top of your QuickBooks or Xero bill.

Does Synder or A2X replace QuickBooks?

No. Both are connectors that need QuickBooks Online or Xero (A2X also posts to NetSuite and Sage) as the destination ledger. If you want one tool instead of two, that is the ERPClaw-shaped path above, where the ledger and the integration are the same product.

Which is better for reconciling Shopify payouts?

A2X, structurally. Its one-entry-per-payout method nets to the bank deposit, so reconciliation is a single match. Synder reconstructs deposits through a clearing account, which works but adds a step, especially in per-transaction mode.

Can I get per-order detail without flooding QuickBooks?

Not really, and that is the structural tension of every connector: detail and GL volume are the same dial. A2X resolves it by summarizing, Synder by posting everything and offering summaries when QBO groans. Keeping per-order detail without a third-party GL means running books where the orders land, which is the design choice behind ERPClaw’s Shopify integration.

## Where to go next

- A2X alternative: the free open-source comparison

- Synder alternative: one tool, not two

- Bookkeep alternative for Shopify accounting

- Free Shopify accounting software in 2026

If you want to see the no-connector path with your own store data, book a quick demo with a co-founder or install ERPClaw and run it beside your current stack for a month. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

## Sources

- A2X, Shopify pricing (tier prices, order caps, store counts, COGS and gift card gates)

- A2X, Shopify accounting page (summary-entry method, destination platforms, customer count claim)

- Synder, Pricing (plan structure and trial terms)

- CostBench, Synder pricing 2026 (per-tier transaction caps, integrations, users, Smart Rules allowances)

- Ottit, A2X vs Synder for Shopify, 2026 (practitioner comparison: GL volume, reconciliation workflow, per-channel subscriptions, client split)

- Intuit, QuickBooks Online pricing (plan prices used in the stack math)

Tagsa2xsyndershopifyaccountingcomparison
Share:XLinkedInHacker News

---

## Adding an Industry to an ERP: a Recipe, Not a Kitchen
URL: https://www.erpclaw.ai/blog/adding-an-industry-is-a-recipe-not-retraining/

# Adding an Industry to an ERP: a Recipe, Not a Kitchen
Why a new vertical in an AI-native ERP is a manifest rather than model training, what a generated tattoo-parlour module actually required, and where the analogy breaks.

Short answer. A trained chef already knows knife work, heat, and timing. Hand them a recipe for a dish they have never cooked and they will make it tonight, because the recipe carries what is specific and the chef carries what is general. That split is the whole argument for how industry verticals should work in an AI-native ERP: the model is the chef, the module manifest is the recipe, and the shared database is the kitchen. Adding an industry means writing one more recipe, not sending the chef back to culinary school. Disclosure: we build ERPClaw and this is how it is built, so weigh the framing accordingly.

## What conventional ERP does instead

Ask a traditional ERP vendor for a vertical they do not have and you are asking for a second kitchen. New tables, new screens, new business logic, new integration surface, and a services engagement to connect it to the accounting that already exists. The work is real, which is why it is quoted in months rather than weeks.

The comparators usually cited for a mid-market implementation are 15.5 months and around $450,000. Both are reported industry survey medians from Panorama’s 2024 research, not controlled measurements of anything, and they should be read as the shape of the market rather than a benchmark. The direction is what matters: a new vertical in that world is a project, not a configuration.

## What the recipe carries

In an AI-native system the split falls differently, because the general knowledge lives in the model and the specific knowledge lives in a manifest.

The chef, meaning the model, already knows how double entry works, what a debit is, why a ledger balances, how a receivable ages, and what a period close means. None of that has to be taught per industry, because none of it changes per industry. A tattoo parlour and a haulage firm keep books the same way.

The recipe, meaning the manifest, carries what is genuinely specific: the entities this industry has that others do not, the actions its operators actually perform, and the handful of business preferences that differ from shop to shop.

The kitchen, meaning the shared database with one writer per table, is what makes the recipe safe to run. A new module reads what it needs and writes only its own tables, so a new vertical cannot corrupt the sales ledger by accident. That ownership rule gets its own treatment in one person holds the pen, and the reason an AI needs deterministic code underneath it is in why an AI cannot be your accounting system on its own.

## The instance, with the number that surprised us

We generated a module for a tattoo parlour. It came out small, a handful of tables and the actions an operator of one actually performs, and its whole generated test suite passed.

The number worth reporting is not any of those. Across three generated modules, grooming, tattoo, and storage, there were 19 human decisions in total, and every single one was a business preference rather than an accounting decision. How long a deposit is held before it is forfeit. Whether a no-show is billed. Whether a storage unit is prorated on exit.

Nobody was asked how double entry works. Nobody was asked which account a deposit posts to, or whether the books should balance. Those questions never surfaced, because they were never industry-specific in the first place. That is the claim in its most concrete form: the decisions that remain are the ones a business owner should be making, and they are the only ones left.

## Where the analogy breaks, and it does

Three limits worth stating plainly, because an analogy that survives every objection is usually hiding something.

A recipe assumes the dish is cookable with the equipment present. A vertical that needs something genuinely absent from the kitchen, a new regulatory filing regime, a physical device integration, a compliance certification, is not a manifest. It is kitchen work, and it costs what kitchen work costs.

A fully passing suite is a statement about the tests, not about the world. It says the generated module does what its specification described. Whether the specification described the industry correctly is a separate question, and the only honest answer is that a practitioner has to look.

Three modules is three. It is enough to show the pattern holds and not enough to characterise its limits. We would rather say that than round it up into a trend.

## Why this is a reuse argument, not a configuration argument

The objection we hear most is that this is just configuration with better marketing. It is a fair challenge and worth answering directly.

Configuration implies switches on a fixed system: the entities exist, and you are turning things on. What happens here is that a module which did not exist is generated, with its own tables and its own actions, against an interface contract the rest of the system already honours. What gets reused is the accounting core and the contract. What varies is the manifest.

Whether that deserves the word reuse or the word generation is a fair argument to have. The distinction that survives either label is the one that matters commercially: conventional ERP answers a new industry by building a second system and connecting it. This answers it by writing one more recipe for a chef who can already cook.

## Where this came from

This is one of the explanations we developed for the IEEE International Conference on Information Reuse and Integration in July 2026, where the reuse question was put to us directly by an audience whose entire field is reuse. It survived that room, which is a better test than surviving our own.

Tagsai-nativeverticalsarchitectureerpreuse
Share:XLinkedInHacker News

---

## Agency Accounting, Project P&L, and Time Billing
URL: https://www.erpclaw.ai/blog/agency-accounting-time-billing-retainers/

# Agency Accounting, Project P&L, and Time Billing
Why your friendly retainer client is secretly losing you $40 an hour, the four metrics every agency owner should know, and how to fix project P&L.

You have a client who is a delight. They pay every invoice within a week. They never argue about scope. They send you holiday cards.

You are also losing about $40 an hour every time your team touches their account. You just do not know it yet.

This is the most common pattern in agency accounting, and almost nobody talks about it honestly. Project P&L for agencies is a mess of three disconnected tools, hours that vanish between Toggl and the invoice, and a year-end chat with the bookkeeper that goes “your books say you made $180K but your bank account says $42K.” If you run a creative shop, dev studio, marketing agency, or consulting firm with 3 to 25 people, this is written for you.

We will walk the four metrics every agency should know, do the math on a real retainer and a real fixed-fee project, then look at what an agency time billing software setup actually needs to do. ERPClaw is the tool we built to solve this, and it is free. But you should understand the problem first.

## The friendly client who is bleeding you

Here is an example. Bruce at Wayne Enterprises is your dream client. He pays you a $5,000 monthly retainer for “ongoing creative support.” Predictable revenue. You love Bruce.

Here is what is actually happening inside that retainer. Your senior designer spends 18 hours a month on Wayne work. Your developer spends 12 hours fixing things Bruce broke on staging. Your project manager spends 8 hours a month on check-ins, weekly status calls, and the Friday “quick favor” that turns into half a day. That is 38 hours a month.

For example, your blended cost (salary, taxes, benefits, software, rent allocated per head) is around $85 an hour. So Bruce costs you 38 times $85, which is $3,230 a month in real cost. You bill him $5,000. Margin looks like $1,770. You feel good.

For example, you also forgot the subcontractor. Bruce wanted a quick illustration last month and you brought in a 1099 freelancer for $600. That is not in your time tracker. You also forgot the Adobe stock licenses you bought specifically for Wayne work, $140 a month. And the project manager undercounted by 6 hours because she stopped logging the Slack DMs after week two.

For example, real cost: $3,230 plus $600 plus $140 plus (6 times $85) which is $510. Total: $4,480. Real margin on Bruce: $520 a month. Or in the bad months, negative.

Your “great client” is paying you a 10% margin while you assume 35%. That is a $40 an hour delta on the team that touches his work. Multiply by 38 hours, multiply by 12 months, and Bruce costs your agency about $18,000 a year in invisible loss compared to what your gut thinks he is worth.

This is not Bruce’s fault. He is doing exactly what the contract says. The fault is that your agency accounting setup cannot tell you the truth about him until it is too late.

## Hours billed are not hours collected

Here is the agency money problem in one sentence: hours billed are not hours collected, hours collected are not project profit, project profit is not annual margin.

Hours billed vs. hours collected. Your team logs 100 hours. You invoice for 100. The client pays for 87 because 13 got marked as “scope clarification” and you swallowed them to keep the relationship warm. Your time tracker says 100. Your bank says 87. Your accountant says nothing because the invoice software does not talk to the time tracker.

Hours collected vs. project profit. Say you collected on 87 hours at $150 per hour, so $13,050. Cost of the team for those 87 hours at $85 blended is $7,395. Plus a $1,200 stock asset purchase, plus $800 to a freelance copywriter. Project profit is $3,655, not $13,050. Most agency owners think of revenue and stop there.

Project profit vs. annual margin. For example, that $3,655 looked great. But across the year, half your projects came in 30% over budget, utilization was 58%, overhead grew faster than billings. Annual margin: 11%. You feel like you are working twice as hard for the same take-home. That is because you are.

The way through this is to measure the four numbers an agency runs on. Most agencies measure zero. The good ones measure two. The great ones measure all four every month.

## The four metrics every agency owner should know

These are the agency equivalent of “revenue, COGS, gross margin, net margin” for a product business. If you can quote your numbers on these four off the top of your head, you are running a real business.

### 1. Utilization rate

Of every hour your team is on the clock, how many were billable to a client? A full-time employee has roughly 2,000 work hours a year. If 1,200 were billable, that is 60% utilization. Healthy is 65 to 75 percent for senior staff, 75 to 85 percent for junior staff. Below 50 percent and you are subsidizing payroll. Above 90 percent and your team is about to quit.

### 2. Realization rate

Of every billable hour you tracked, how many made it onto an invoice the client paid? This is where the “I will eat that scope creep” tax shows up. If you tracked 1,200 billable hours but only collected on 1,020, your realization rate is 85 percent. Below 80 percent and you are giving away serious money. The fix is usually not “bill harder,” it is “stop scoping fixed-fee work that doubles in size.”

### 3. Collection rate

Of every dollar you invoiced, how many arrived in your bank account within 60 days? Below 92 percent and you have a client filtering problem. Bad clients are not clients, they are unpaid loans.

### 4. Project margin

Per project: revenue minus all direct cost (team hours at blended rate, plus subs, plus pass-through). Healthy is 50 to 65 percent for creative and dev, 35 to 50 percent for low-touch retainer, 25 percent or below and you are losing money once overhead is added. The trick is calculating it per project, not for the agency overall. The agency-overall number hides the Bruces.

If you measure those four monthly, you will catch a Bruce-shaped problem in eight weeks instead of eight months.

## Worked example: Wayne Enterprises plus Stark Industries

Let us run the math on a real month at a hypothetical 8-person agency.

Client A: Wayne Enterprises. Suppose a $5,000 monthly retainer.

- Senior designer: 18 hours, developer: 12 hours, PM: 8 hours

- Subcontractor (illustrator): $600. Stock assets: $140.

Total hours: 38. Labor cost at $85 blended: $3,230. Subs and pass-through: $740. Total cost: $3,970. Revenue: $5,000. Project margin: $1,030, or 21 percent. Below the healthy 35 percent floor for a retainer. Wayne is a watch-list client. One more “quick favor” and he goes red.

Client B: Stark Industries. Suppose a fixed-fee project bid at $18,000 for a brand refresh. You estimated 100 hours of senior design plus 20 hours of PM. Estimated cost: 120 times $85 equals $10,200. Estimated margin: 43 percent. Clean.

Suppose this is what actually happened: Tony kept asking for “one more direction.” Pepper signed off then unsigned off. By the time you delivered, you had spent 145 hours of senior design and 20 hours of PM. Plus a brand strategist sub for $2,400 because you needed positioning work you had not scoped.

For example, real cost: 165 hours times $85 equals $14,025, plus the $2,400 sub equals $16,425. Revenue: $18,000. Project margin: $1,575, or 8.75 percent. You billed Tony for the full $18K. He paid on time. Your bank account smiles. Your project P&L is screaming.

Suppose you knew that number on the day it crossed 25 percent margin (around hour 130), you could have called Tony and renegotiated. You would have had the data to say “we are 30 percent over the original scope, here is the change order.” Instead you found out at the post-mortem, three months too late. This is the difference between agency accounting that runs your business and agency accounting that just files your taxes.

## How agencies typically track this badly

Walk into any 10-person agency and ask “what tools do you use for finance and ops?” Here is the answer you get nine times out of ten:

- Toggl or Harvest for time tracking. Reports get exported to CSV monthly.

- A spreadsheet to map time entries to projects and clients. Lives on the founder’s laptop. Has a tab called “do not delete.”

- FreshBooks, Bonsai, or QuickBooks Online for invoicing. Has no idea what hours those invoices represent.

- A bookkeeper who reconciles the bank to the invoice software once a quarter. Sends a P&L PDF that nobody reads.

Four tools. Four different sources of truth. Zero of them know what the other three know. The founder spends one Saturday a month duct-taping a CSV from tool 1 into the spreadsheet from tool 2 to figure out what to invoice in tool 3.

Project P&L is theoretically possible in this setup. In practice nobody does it because it would take 6 hours per project per month and the founder has client work to do. So you find out about the Wayne problem in March when you are doing taxes. You find out about the Stark margin compression three months after delivery. You hire a new designer because “we feel busy” without knowing if your team is at 55 percent utilization or 85 percent.

This is not an “agency owners are bad at money” problem. It is a tooling problem. The tools exist to do each piece, but no consumer-grade agency tool puts them in one database where the math works.

## What an agency-friendly accounting setup looks like

Forget brand names for a minute. Here is what a sane agency setup does as one connected system:

- A team member logs hours against a client and project. One screen. Ten seconds.

- Hours roll up to the project automatically. Project page shows hours logged, hours remaining against scope, blended cost, and current margin.

- At month end or milestone, an invoice is generated from the hours and any flat fees. The invoice already knows which hours, which project, which client. No copy-pasting.

- The client pays via Stripe (or ACH, or wire). Payment lands, invoice marked paid automatically.

- The bank deposit reconciles to the invoice automatically. No bookkeeper matching line items.

- The P&L updates in real time, per client and per project. Ask “what was my margin on Bruce last month” and get an answer in two seconds.

- Subcontractor invoices and pass-through expenses attach to the project. Margin includes them automatically.

That is the whole shape. It is not exotic. It is just that no single $30-a-month tool does all seven connected to each other. You either pay $400 a month for an enterprise system, or you live with the four-tool tape-and-string approach.

## How ERPClaw does this in one tool

ERPClaw is an open-source, AI-native ERP we built specifically because the four-tool problem was driving every agency owner we talked to slightly insane. The agency setup is one of the prebuilt configurations on the agency vertical page.

Here is what runs in one SQLite database when you install it:

- Projects module. Each project has a budget (hours and dollars), assigned team, and a live P&L view that updates as hours come in.

- HR / time tracking. Team members log hours against a project. The hours flow straight into project cost at the configured blended rate.

- Selling and invoicing. Generate retainer invoices on a schedule, milestone invoices when a project hits a phase, or time-and-materials invoices that pull hours straight from the time tracker. All from the same database.

- Stripe integration. Client pays the invoice via Stripe. Payment lands, invoice gets marked paid, journal entries hit the accounting books in the same transaction. No manual reconciliation.

- Accounting and GL. Real double-entry under the hood, so a CPA takes it seriously, but you never have to look at it. You see “client paid you $5,000, here is your updated margin.”

- Reports. Project margin, client profitability, utilization rate, realization rate, collection rate. The four metrics are built in. Because ERPClaw is AI-native rather than AI-decorated, you ask the assistant “what is my margin on Wayne Enterprises this quarter” in plain English and get the answer, instead of clicking through a reports menu in a tool that bolted a chat sidebar onto a 2010 form UI.

Total cost: $0. Runs on your laptop, a Mac mini in the closet, or a $5 VPS. Data lives in one file you back up to Dropbox. No per-seat pricing because there is no SaaS model. The full feature list is here and the pricing page is short because the answer is “free, forever, all included modules.”

The pitch is not “ERPClaw is fancier than your current tools.” The pitch is “ERPClaw is the single database those four tools were pretending to be.”

## When your agency is too small for ERPClaw

I will talk you out of installing it if you do not need it.

If you are a 1 to 3 person consultancy with a handful of retainer clients, no subcontractors, no inventory, and you do not care about per-project margin because every project is basically you, FreshBooks at $19 a month is genuinely fine. So is Bonsai. So is a Notion board plus Stripe Invoicing. The double-entry rigor that ERPClaw provides is overkill for a one-person shop. The FreshBooks comparison walks through where the line is.

You start outgrowing the four-tool setup at roughly: 4+ team members logging hours, 8+ active client engagements, a mix of retainer and project work, meaningful subcontractor spend, or a founder who has stopped sleeping because the spreadsheet is wrong somewhere. Two or more of those and the four-tool tax is costing you more than the migration would.

## Outbound reading

Two pieces I send to founders who want to go deeper on the agency profit math: the long-running discussion of utilization and pricing on Smashing Magazine and the Bureau of Digital community writeups. Both more honest than most agency-software vendor blogs.

## FAQ

### What is the best agency accounting software for a 10-person creative shop?

The test is whether the tool gives you per-project P&L without copy-pasting from a time tracker. If you are willing to pay, Productive and Accelo do this. If you want free and self-hosted, ERPClaw is the open-source option. If you only need invoicing and you do not care about project margin, FreshBooks is enough.

### How do you calculate project P&L for an agency?

Per project: revenue collected minus labor cost (hours logged times blended cost rate) minus subcontractor invoices tied to the project minus pass-through expenses (stock assets, ad spend, hosting). The result divided by revenue gives margin percentage. Do this monthly, not at project end.

### What is a healthy utilization rate at an agency?

Senior staff: 65 to 75 percent. Junior staff: 75 to 85 percent. Below 50 percent across the team and the agency is unprofitable. Above 90 percent and your team is two months from burning out. Those bands are the rough goalposts in the consulting agency project margin literature.

### How is an agency ERP different from project management software?

Project management software (Asana, Monday, ClickUp) tracks tasks and timelines. An agency ERP tracks money: hours, costs, invoices, payments, margin, reconciled to the bank. Most agencies use both. The mistake is using project management as if it were accounting, which is how invoices get sent late and projects come in over budget without anyone noticing.

### Can ERPClaw bill both retainers and hourly projects for the same client?

Yes. Most agency tools assume one billing model per client. ERPClaw lets a single client have a retainer, several active hourly projects, and a fixed-fee milestone project at once, all billed from the same chart of accounts. The agency configuration is set up that way out of the box.

### What about 1099 subcontractors and year-end forms?

Subcontractor invoices attach to projects, so their cost rolls into project margin automatically. Vendor records track the W-9 info for each 1099 contractor. At year end, ERPClaw generates the totals you need for 1099-NEC filings without a separate tool.

## The boring truth

Most agencies are profitable on paper and miserable in practice because the four-tool stack is hiding where the money actually goes. Bruce is not a great client until you have measured him. Tony is not a profitable project until you have priced the change order.

The fix is one tool, one database, four metrics, every month. If you want to do that with FreshBooks and a spreadsheet, you can. If you want one open-source tool that a 10-person agency can actually run, the agency configuration of ERPClaw is free and installs in about ten minutes. Either way, find out what Bruce is really worth before another year goes by.

---

## AI-Decorated vs AI-Native: Why Most AI Features Lose
URL: https://www.erpclaw.ai/blog/ai-decorated-vs-ai-native-software/

# AI-Decorated vs AI-Native: Why Most AI Features Lose
AI-decorated tools bolt a chatbot onto 2015 software and charge a new fee. AI-native software rebuilds the architecture. One of these wins. Here is why.

In 2008, almost every company with a website added a “mobile-friendly” version. They shrunk the columns, hid the sidebars, called it done. A smaller group threw the website away and built something different: a touch-first, sensor-aware, location-rich application that assumed the phone as the primary surface. The companies that decorated their desktop product with a mobile skin became case studies. The companies that rebuilt around mobile primitives became Uber, Instagram, WhatsApp, and most of the consumer internet.

We are repeating the exercise with AI in 2026, and most software companies are once again on the wrong side of the line. The distinction that matters this year is AI-decorated vs AI-native software. This post is the framework I use to tell them apart, the reason one of them is going to lose, and what it means if you are buying software, building it, or funding it. For the ERP cut of this argument see our pillar on AI-native ERP; for the accounting cut see AI-native accounting, and for a survey of how the whole accounting field is splitting along this line, the state of AI in accounting in 2026.

## What AI-decorated software actually is

AI-decorated software is the dominant pattern of 2024 and 2025. A vendor with a product that shipped in 2012 adds a chat sidebar, trains a small model on their help docs, puts a sparkle icon next to a few buttons, announces “AI” in the product name, and adds a fee on top of the existing subscription. The architecture, the data model, the pricing model, and the team that maintains the codebase are unchanged. The AI is a feature. The product is the same product.

A short list from the last 24 months:

- Salesforce Einstein and Agentforce. A chat layer over a 1999 data model, priced as an add-on. The CRM is still the CRM.

- QuickBooks Intuit Assist. A summarization and Q&A layer over a 1983 general ledger product, with a separate price tag.

- Microsoft 365 Copilot. A useful assistant inside Word and Excel, sold as a roughly $30 per seat per month add-on (reported enterprise add-on price, 2026) on top of the existing subscription.

- Notion AI, Slack AI, Zoom AI Companion, HubSpot Breeze. Each one a chat-and-summary layer welded onto a product specified before transformer models existed.

These are not bad products. They are decorated, not native, and that distinction will decide which of them survive the next price war. The defining property of an AI-decorated product is that you could remove the AI and the product would still work the same way it did before. The chat sidebar is detachable. The architecture would not notice.

## What AI-native software actually is

AI-native software is the inverse. The product is architected around AI as the primary interface, the primary implementation surface, or both. Remove the AI and the product collapses, because the old workflow was never built.

The clearest examples today:

- Cursor and Claude Code for software engineering. The editor exists to mediate between the developer and the model. The file tree, multi-file edit semantics, context window management, diff review surface: every primitive is shaped by what the model needs.

- Perplexity for search. The retrieval, citation rendering, follow-up question UI, and answer-first layout assume an LLM at the center. Google’s “AI Overviews” are decoration over the same ten blue links Perplexity threw out.

- Granola for meeting notes. The capture, silent listening, structured output, and post-meeting editor are designed around a model with the full transcript in context. Otter bolted transcription onto a 2015 product; Granola was specified around the model.

- ERPClaw for ERP. The system our team built. Architectural detail below.

The defining property of an AI-native product is that the model is load-bearing. Take it out and there is no product underneath.

## The economic argument: AI-decorated cannot win on price

Here is the part founders and investors most need to internalize.

An AI-decorated vendor has the cost structure of the original 2010 to 2020 era product, plus the cost of the AI added on top. They pay for the engineering team that built and maintains the legacy codebase, the data centers, the support org, the sales motion, and the per-seat margins their board expects. The AI is an add-on, both architecturally and on the invoice. That is why every product listed above charges an extra monthly fee for AI: the cost base demanded it.

An AI-native vendor that started in 2024 or later has a structurally different cost base. The codebase is smaller because the model writes most of the mechanical code. The team is smaller because coordination overhead drops with team size. The pricing is not anchored to a per-seat assumption inherited from the on-prem era.

The result is the asymmetry that always shows up in technology shifts. The new entrant can charge half of what the incumbent charges and still have better margins, or give the core product away (the path I picked) and monetize the parts the model cannot do for you. Either way, the AI-decorated tool you are paying a per-seat subscription for in 2026 will look badly priced in 2027 and indefensible by 2028.

This is the Kodak failure mode, almost exactly. Kodak invented digital photography. What they could not do was abandon the cost structure of their film business fast enough to compete with a business that was digital from day one. The same dynamic is playing out across every category of business software.

## The 80/15/5 theory of where AI eats first

Why is the AI-native cost base so much lower? Here is the rough split I use, drawn from building ERPClaw rather than from any survey, call it the 80/15/5 split:

- Roughly 80 percent is mechanical work. Schema scaffolding, CRUD endpoints, audit logging, parameter validation, list and get APIs, the boring form code every business application needs. Repetitive, well specified, identical across products. AI does it well when given a precise spec.

- Roughly 15 percent is pattern-matching work. Cross-module integration, report templates, common workflow shapes, GL posting patterns. AI does this well with the right examples and a checklist of invariants. A human catches the cross-cutting surprises.

- The last 5 percent or so is human judgment. Domain edge cases, regulatory nuance, the question the AI did not know to ask, the business decision that requires taste. Irreducibly human.

These are my estimates, not a measured benchmark, but the shape holds in my experience. A traditional vendor with a 200 person engineering team spends most of its salary line on that mechanical majority. An AI-native vendor automates that and reorganizes around the small slice of judgment work that matters. The new tools are not written by smarter people; they are written by people who do not have to write the boring code at all.

This is also why you cannot retrofit AI-native. You cannot fire the bulk of the team that maintains the legacy codebase, because somebody still has to keep it running. You cannot rewrite a million-line codebase from scratch on a quarterly earnings cycle. Decorated vendors get stuck paying for both stacks in parallel. The math does not work.

## How to tell if a product is actually AI-native

Every software company on earth now claims to be “AI-first” or “AI-native.” Most are decorated. The test I run:

1. Is the AI load-bearing or detachable? If you removed the AI feature tomorrow, would users notice? In a decorated product they would shrug and use the old menus. In a native product the workflow stops working.

2. Is the pricing additive or unified? Decorated vendors charge a base fee and an “AI” fee on top, because the AI is a discrete cost line they need to recover. Native vendors price the whole product as one thing, because there is no separate “non-AI” version.

3. Is the team small or 200 people? If the vendor has thousands of engineers, that cost base will show up in the price, regardless of whether the demo had a sparkle icon.

4. Was the data model designed before or after 2023? Schemas designed before LLMs were practical were optimized for human form-filling and SQL reporting. AI-native schemas are optimized for model context: metadata-rich, self-describing, queryable by an agent without a translation layer. If the API looks like 2008 REST with a chat endpoint glued to the side, it is decorated.

5. Could the codebase be regenerated from a spec? The deepest test. In an AI-native product the spec is the asset and the code is a derived artifact; large parts of the system can be regenerated in a weekend if a better architecture appears. In a decorated product the code is the asset because somebody paid millions to write it, and that asset is exactly what they cannot afford to throw away. The sunk cost is now a liability.

Four out of five is plausibly native. One or two is decoration with a press release.

## ERPClaw as a worked example

A concrete proof point: ERPClaw, the AI-native ERP our team has built over the last fifteen months. ERP is a category where the AI-decorated pattern is most visible (NetSuite, Sage, Microsoft Dynamics have all shipped a chat sidebar and a price increase in the last 18 months).

ERPClaw is an open source ERP that covers the full back office (sales, purchasing, stock, manufacturing, HR, payroll, projects, and more) across 14 industry verticals, with a deep, scriptable action and schema surface across the whole system. It would not exist as a hand-written codebase in any reasonable timeframe. It exists because the architecture was designed around three primitives that only make sense in an AI-native world.

First, the spec is the source of truth, the code is regenerable. Every action has a YAML-fronted SKILL.md entry that defines parameters, return shape, and invariants. The implementation, the contract tests, the dashboard forms, and the docs are all generated from that spec. A traditional ERP hand-writes roughly 150 lines of UI code per action; across a system this broad that is hundreds of thousands of lines that simply do not exist in this codebase.

Second, a constitutional rules engine catches AI mistakes the AI cannot catch itself. A set of machine-readable financial laws (no floats for money, GL is immutable, every submit is a single transaction, trial balance balanced after every test) run as executable assertions across the codebase. A module that violates an article cannot ship. That is what makes it safe to extend the system at AI speed without waking up to an expensive audit finding.

Third, the pricing reflects the cost base, not the legacy anchor. The open source edition is free. Not freemium, free. The marginal cost of generating a new module is hours of model time, not weeks of engineering salary (adding an industry is a recipe, not retraining shows what that looks like when a new vertical ships). Decorated incumbents cannot match that arithmetic.

I am not telling you to install ERPClaw. I am telling you what an AI-native build looks like in a category everybody else is decorating. The same shape will appear in every other commodity software category over the next 36 months. CRM, project management, HR, helpdesk, billing, BI: commodity problems with public business rules, dominated by decorated incumbents, about to be re-priced by AI-native entrants. More on how this came together in the build retrospective and the founder origin post.

## What this means if you are a buyer

Three implications if you are signing a software contract in 2026.

Audit the AI line items on every renewal. A separate fee for AI features on top of the base subscription is a tell. It signals an architecture and a cost base that is going to be undercut. Negotiate the AI line item to zero this year, or start evaluating the native alternative before the price comparison gets ugly in 2027.

Stop signing five-year contracts in commodity categories. ERP, CRM, HR, helpdesk, BI: every one of these is going to look different in 36 months. A long contract with a decorated vendor is a bet that nothing changes. Annual contracts with break clauses are sane.

Run the five-question test on every AI-marketed product you evaluate. If a vendor cannot articulate why their AI is load-bearing, why their pricing is unified, and how their codebase could be regenerated, you are looking at decoration. Use a decorated product if it solves your problem today. Just do not pay AI-native prices for AI-decorated work, and do not assume the vendor will catch up.

## What this means if you are a builder

Two implications for founders and CTOs.

You cannot retrofit AI-native onto a legacy codebase. I have watched several teams try, and the failure mode is always the same: the new AI-native module ships, it impresses the demo audience, and the maintenance cost of running both stacks in parallel quietly destroys the engineering org over 18 months. If your category is being attacked by an AI-native entrant, the only response that works is to start a parallel codebase from scratch with a small team and a precise spec, and accept that the legacy revenue is going to decline while you build it. That move worked for Adobe (Creative Cloud) and Microsoft (Azure). It did not work for Kodak.

The spec is the moat now, not the code. If the model is going to write the code, the differentiated artifact is the specification, the constitutional invariants, and the test suite that catches model mistakes. Spend your engineering hours on those. The code is a derived asset and increasingly cheap. The spec is where taste, domain knowledge, and judgment live. The a16z piece on AI-native applications makes a related point in the consumer context: the unit of differentiation is moving up the stack, away from implementation and toward design.

## A closing word on the vocabulary

“AI-native” and “AI-decorated” will be co-opted by marketing departments within twelve months, the way “cloud-native” lost meaning between 2014 and 2018. The vocabulary will rot. The underlying distinction will not. Whatever we call the two camps in 2028, the decorated cohort will still be losing share to the native one.

The question to ask of any AI feature right now is not “does this look impressive in a demo.” It is is the AI load-bearing, and is the price reflecting an AI-native cost base? For what the answers look like in ERP, the features overview and the pricing page are the exhibits.

## FAQ

What is AI-native software in one sentence?

Software whose architecture, pricing, and team shape were designed around AI as a load-bearing primitive, such that removing the AI would collapse the product rather than degrade a feature.

Is AI-first the same as AI-native?

No. “AI-first” is a marketing claim about strategic priority. AI-native is an architectural claim about how the product is built. Plenty of AI-first companies ship AI-decorated products, because the priority did not survive contact with the legacy codebase.

Is your AI feature actually native?

Apply the five-question test: load-bearing or detachable, unified or additive pricing, small team or 200 people, data model designed for AI context or for SQL reporting, codebase regenerable from a spec or not. Four out of five is plausibly native.

Will AI-decorated vendors ever catch up?

Some will, the way some on-prem vendors successfully became cloud companies. Most will not, because running both stacks in parallel is structurally unsustainable. Watch for the ones that announce a “next-generation platform” and price it separately from the existing product. That is the shape of a serious attempt.

Why is open source the natural home for AI-native software?

The marginal cost of generating new modules is small, the spec is the moat rather than the code, and the cost base does not require per-seat pricing to recover engineering salaries. Open source AI-native systems will eat commodity SaaS the way Linux ate proprietary Unix. The vendors who survive will sell domain expertise, regulatory compliance, distribution, and trust, not code.

Where does ERPClaw fit in this taxonomy?

ERPClaw is an AI-native ERP, open source, built on the spec-first methodology. It passes the five-question test by design. A worked example of an AI-native build in a category incumbents are decorating.

What should I do this quarter if I am a buyer?

Inventory your contracts. Mark the ones with separate AI line items and the ones longer than 24 months. Start a quiet evaluation of the native alternative in your top three categories.

---

## AI-Native vs the AI Layer: Two Ways to Put AI in Your Books
URL: https://www.erpclaw.ai/blog/ai-native-vs-the-ai-layer/

# AI-Native vs the AI Layer: Two Ways to Put AI in Your Books
AI on top of your books, or AI-native books? The two 2026 patterns compared with named vendors, real prices, and what each means for owners.

Two kinds of companies are selling “AI for your business systems” in 2026, and they are not selling the same thing. It is worth naming the pattern precisely, because the two look identical in a demo and could not be more different three years into owning one.

The first kind puts AI on top of a system of record it does not control. Call it the AI layer. The second rebuilds the system of record itself so the AI is the interface, not an attachment. That is AI-native. ERPClaw is in the second camp, so read this knowing where we stand; every claim about the companies named is from their own public materials, with dates.

## The layer pattern, in their own words

The layer companies are explicit about the pattern, and honest about it.

Billow AI Labs (Y Combinator summer 2026) describes “zero software for customers to use, just integrate and we’ll close your books for you.” Its system connects to NetSuite, QuickBooks, Xero, and a long list of finance tools, reads the books, posts journal entries, and routes exceptions to humans. The incumbent ERP stays, and so does its bill.

Definite (definite.app) pitches “the AI-native data platform”: it centralizes your billing, accounting, and revenue data into its own store and answers questions in plain English. It is candid that it reads from your systems and does not write back. Pricing runs from a free tier to $250 per month and up on their published tiers (checked August 2026).

Spaceflow (spaceflow.tech, also Y Combinator summer 2026) runs AI agents for procurement “inside the systems you already run,” with SAP named first. Their framing is disarming: “no re-implementation, no data-cleansing megaproject.” Every agent action is “proposed, logged, and approved by your team,” and the ERP “stays the system of record.”

Three different products, one architecture: the system of record is somebody else’s, so the AI must sit beside it, copy from it, or ask permission to write into it.

## What the layer can never fix

None of this is a scam; layering is a rational response to systems too entrenched to replace. But the architecture has a floor it cannot dig below.

The underlying bill survives. Every layer presumes the ERP subscription underneath keeps running. Whatever the layer costs, it is additive.

The layer sees a copy or borrows a pen. A read-only layer answers questions from a cache of your books, and a writing layer is a third party posting into your ledger under supervision. Either way, the thing being analyzed and the thing being true are held apart by an integration.

Growth means another layer. The close layer does not do analytics; the analytics layer does not do procurement; the procurement agents do not do the close. Each solves one slice, each meters separately, and the stack of AI helpers starts to look like the module pricing sheet it was supposed to replace.

## The AI-native alternative

AI-native means the plain-language request is the system’s primary interface, and everything under it was built for that. In ERPClaw: you say “invoice Acme for the March retainer, 4,500 dollars, due in 30 days,” the AI resolves what you meant, deterministic accounting code posts the entry, and an invariant engine checks the books after every posting. The AI never improvises a debit. It decides what you meant, never what the books say.

Do that at the ledger itself and the layer’s reasons to exist fall away one by one. The close is not a monthly crisis, because the books were never allowed to drift. The analytics need no copy, because the questions run against the ledger they are about. And there is no second bill, because ERPClaw is open source, self-hosted, and $0 forever.

The AI-native accounting startups, Campfire, Rillet, DualEntry, Puzzle, made the same architectural bet we did, and their funding says the market believes it: Rillet raised a $70M Series B in August 2025, Campfire a $65M Series B in October 2025, DualEntry a $90M Series A that same month, all on AI-native theses. Where they differ from ERPClaw is not architecture but custody and scope: all four are closed SaaS, most with unpublished pricing, and their coverage is the finance suite, with inventory, purchasing, and payroll handled by integrations or absent. We compare against each honestly on their own pages: Campfire, Rillet, DualEntry, Puzzle, Billow.

## The question to ask any vendor

One question separates the patterns in a demo: “When your AI acts, whose system holds the result, and who checks it?”

If the answer involves another company’s database, a sync, or a human reviewing a third party’s journal entries, you are buying a layer. Sometimes that is the right purchase; a team locked into NetSuite for five more years should absolutely make the five years bearable.

But if the ERP decision is still open, the order matters. Decide the ledger first. A ledger that is AI-native, open source, and yours makes most of the layers unnecessary, and it is the one part of the stack you should never rent.

The whole-stack version of this comparison, slice by slice, is at ERPClaw vs the AI-ERP stack you’d otherwise assemble.

Tagsai-nativeai-layererparchitecturepositioning
Share:XLinkedInHacker News

---

## Bookkeep Alternative: Daily Summaries Cost You Later
URL: https://www.erpclaw.ai/blog/bookkeep-alternative-shopify-accounting/

# Bookkeep Alternative: Daily Summaries Cost You Later
Bookkeep is a cheap Bookkeep alternative to A2X, but daily summaries cost you at audit time. ERPClaw posts every Shopify order to your books, free.

It is the first week of April. Suppose you open Bookkeep to send your March numbers to your CPA. The export looks great, one tidy daily summary line per day, thirty one neat rows for the month. You forward it. Twenty minutes later your CPA replies, “Where are the individual transactions? I need to see the $89 refund on the 14th. Bank shows it but the books just say one daily total.”

That is the moment most stores realize the Bookkeep alternative question is not really about price. It is about what gets recorded in your books and what does not. If you are paying $19 to $199 a month (as of June 2026) for Bookkeep and wondering whether a cheaper or better Bookkeep alternative exists, this post is for you. Short version: the better alternative is also free, runs on your own machine, and posts every order separately so the $89 refund never disappears into a $14,000 daily total.

That alternative is ERPClaw. Our team built it because Shopify store owners were paying monthly bills for software that aggregated their data into shapes their accountants could not use. Below is the honest comparison, including where Bookkeep is genuinely fine.

One framing point before the post starts. Bookkeep is AI-decorated software (a connector built before AI was practical, with assistant features added on top). ERPClaw is the only AI-native option in this category, which means the AI assistant is the primary interface, not a sparkle icon glued to the right rail. The difference matters because you ask the assistant questions like “show me every refund booked under the wrong account this quarter” in plain English, instead of clicking through three menus to find the report. The full argument is in AI-decorated vs AI-native software.

## What Bookkeep does well

Before the criticism, the credit. Bookkeep earned its customer base by being a real product:

- It is cheap. Their entry tier on the Bookkeep pricing page starts well below A2X. For a small store doing under 200 orders a month, the bill is real but small.

- Setup is fast. Connect Shopify, connect QuickBooks Online or Xero, pick your chart of accounts, done. Most stores are live in under an hour.

- It auto-posts. You do not have to click anything every morning. The daily summary lands in QuickBooks while you sleep.

- Multi-channel. One subscription handles Shopify, Amazon, Square, and a few others. Owners with two or three sales channels get reasonable value.

- Customer support is responsive. I have heard this from multiple store owners. They pick up the phone.

If your store is small, your CPA only ever looks at the month-end totals, and you never get audited, Bookkeep is fine. Genuinely fine. Skip to the “When Bookkeep is fine” section near the bottom and save yourself the read.

## The daily summary trade-off, in one sentence

Bookkeep is cheap precisely because it gives you summaries, not detail. That is the trade.

Here is what that means in practice. Suppose on March 14th your store does $14,000 in revenue across 87 orders. One of those orders, order number 4521, is a $312 sale that gets refunded the same day for $89 (partial refund, customer kept one of the items). With Bookkeep on the daily summary plan, your QuickBooks ledger gets a single line for March 14th that looks something like this:

- March 14: Sales $14,000, Refunds ($89), Net Revenue $13,911

The $89 refund exists in the total. It does not exist as its own entry tied to order 4521. When your bank statement six months later shows a $89 outflow on March 14th and your CPA asks “what was that,” you have to log into Shopify, find order 4521, screenshot the refund, and email it over. Multiply that by every refund and dispute for the year and you have hours of CPA work that bills back to you at $150 an hour.

Per-transaction accounting puts each of those 87 orders in your books as its own entry. Refund 4521 is its own line. The bank match is automatic. The audit trail is one click.

## What auditors actually want

I have sat through enough year-end reviews and one tax audit to know how this plays out. An auditor or a CPA doing a clean-up review wants three things:

- Every line in the bank statement matches a line in the books. Not a summary that nets to the bank deposit. A line per transaction.

- Every refund traces to the original sale. They want to click from the refund to the invoice it reverses.

- Sales tax owed reconciles to the orders that generated it. State by state, jurisdiction by jurisdiction. Daily summaries make this messy because tax gets aggregated before it gets reported.

Daily summary tools can answer all three of these eventually, by going back to Shopify, exporting per-order data, and stitching it together. That is work. The whole point of buying accounting automation is to not do that work.

Per-transaction tools answer all three by clicking on a line in the books. That is the structural difference.

## What ERPClaw does instead

ERPClaw is an open source ERP that ships a Shopify integration as one of its included modules. The Shopify integration is shipped and installable today; the docs cover setup. It pairs to your store with a six character code, and it posts every order, refund, payout, and dispute as its own entry in your books. Free, forever, open source.

The four ways it differs from Bookkeep:

Per-transaction always. Every Shopify order becomes a real sales invoice with line items, taxes, shipping, discounts, and the full Shopify order details attached. Refunds become credit notes pointed at the original invoice. The $89 refund on order 4521 has its own line, every time. No daily summary mode. No paid upgrade required to see the detail.

No QuickBooks required. Bookkeep is a connector. It moves data from Shopify into QuickBooks Online or Xero. You still need the QuickBooks subscription underneath, which is another $38 to $275 a month (as of June 2026). ERPClaw is the books itself. Your sales invoices, your bank reconciliation, your trial balance, your profit and loss, all live inside ERPClaw. You can drop QuickBooks entirely.

Free, not freemium. There is no $20 tier. No $50 tier. No premium upgrade locking the per-transaction view behind a paywall. The full product is free, including the per-transaction posting, multi-warehouse stock costs, gift card deferred revenue handling, and the other included modules. The business model is documented at /pricing.

More than just sales sync. Bookkeep aggregates sales channel data and pushes it into accounting software. ERPClaw also handles your stock, your purchase orders, your vendor bills, your HR, your payroll, and your manufacturing if you have any of those. One app instead of five.

## Side by side

| Feature
| Bookkeep
| ERPClaw

| Price
| $19 to $199 a month
| $0 forever (open source license)

| Granularity
| Daily summaries
| Per-transaction always

| Destination
| QuickBooks / Xero
| ERPClaw (the ERP itself)

| Sales channel coverage
| Shopify, Amazon, others
| Shopify in v1, more coming

| Per-transaction journal entries
| Higher tier only
| Always

| Cost-of-goods tracking
| Limited
| Multi-warehouse, FIFO

| Gift card deferred revenue
| Yes
| Yes

| Self-hosted
| No
| Yes

| Open source
| Proprietary
| Open source (GPL v3)

| Architecture
| AI-decorated (chat sidebar bolted on)
| AI-native (assistant is the primary interface)

| Database
| Their cloud
| SQLite or PostgreSQL via PyPika, on your machine

| Full ERP modules (HR, manufacturing)
| No
| Yes

| API access
| Limited
| Full (CLI + web)

| Needs QuickBooks or Xero
| Yes
| No

The full live comparison is at /compare/bookkeep and gets updated when either product changes.

## When Bookkeep is fine

I want to be fair about this. Bookkeep is the right call in three scenarios:

- You are a one person shop doing under 5 orders a day. Your CPA looks at the monthly P&L, signs off, and moves on. You will never get audited. The simplicity is worth $19 a month.

- Your CPA only ever looks at month-end totals. Some bookkeepers genuinely do not need transaction level detail because they reconcile against bank statements at a summary level. If that is your setup and your CPA is happy, do not break what works.

- You are scared of self-hosting. ERPClaw runs on your own laptop or a low-cost VPS. If that sentence makes you nervous and you would rather pay someone to run software for you, Bookkeep is one option, or have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

If none of those three describe you, the per-transaction route saves you money on the right axis (your CPA bill, your audit risk, your time at tax season) while costing nothing on the wrong axis (the monthly software bill).

## Migrating from Bookkeep, the short version

The full migration walkthrough is being written and will live at `/docs/migration/from-bookkeep` when ready. The shape of it:

Step 1. Pick a switch over date. First day of a new accounting period is cleanest. Most stores pick the start of a month or quarter.

Step 2. Install ERPClaw. Five minutes on your own machine. The full setup is at /docs/shopify/install-walkthrough/.

Step 3. Pair the Shopify integration. The integration is shipped and installable today; the feature tour shows what it does and the docs cover setup. The embedded admin shows a six character pairing code. Run the pair command on your ERPClaw box (the install walkthrough has the exact syntax). There is also a self-hosted custom-app token flow if you prefer that route.

Step 4. Pull in your Shopify history. ERPClaw pulls every order, refund, payout, and dispute from Shopify going back as far as you ask. Most stores pull the last 12 months. Each transaction lands in your books with its original date and full detail.

Step 5. Cross check the books. Compare ERPClaw’s trial balance on the cutover date against the daily summaries Bookkeep handed to QuickBooks. They should match within rounding. If they do not, the per-transaction view in ERPClaw shows you exactly which order or refund is the difference.

Step 6. Cancel Bookkeep and decide on QuickBooks. Save a copy of your Bookkeep mappings as a paper trail for your CPA. Most stores drop QuickBooks too once ERPClaw is the books. Some keep QuickBooks running for a quarter to make their accountant comfortable, then sunset it.

A long Saturday afternoon, max. If you have under 12 months of history, closer to two hours.

## FAQ

Is ERPClaw really free, or is there a Bookkeep alternative tier I should know about?

Really free. Open source. No paid tier. No “pro” upgrade. The full product, including per-transaction posting, multi-warehouse stock costs, and the rest of the ERP modules, is free forever. We make money on ERPClaw Cloud Managed (our hosted offering, onboarding by request through /pricing/), industry consulting, and a future marketplace for third-party add-ons. None of those lock the core product.

Does ERPClaw need QuickBooks underneath like Bookkeep does?

No. ERPClaw is the books. Bookkeep is a sync layer that needs QuickBooks Online or Xero underneath as the actual books of record. ERPClaw replaces both. You can keep QuickBooks running in parallel during a switch over period if your CPA wants the comfort, then drop it.

Will posting every order instead of a daily summary slow my Shopify store down?

No. The Shopify webhook that fires when an order is placed gets booked in ERPClaw within a few seconds, on your own server. A high volume store doing 10,000 orders a month will see 10,000 sales invoices in ERPClaw at month end. SQLite handles that comfortably and reports run in milliseconds. The only slow part is the first history pull during migration, which takes a few minutes per 1,000 orders.

My CPA insists on QuickBooks. Can I still use ERPClaw?

Yes. ERPClaw can push a one way summary export to QuickBooks Online for the period your CPA wants to keep working there. You get per-transaction detail on your side, your CPA gets the QuickBooks view they like, and you sunset the QuickBooks export whenever they are ready to move.

How is ERPClaw different from A2X, since A2X also does per-transaction?

A2X is a paid alternative ($29 to $229 a month, as of June 2026) that posts to QuickBooks or Xero. ERPClaw is free, runs on your own machine, and is the books itself. Full A2X comparison at /compare/a2x.

What about gift cards and deferred revenue?

Both Bookkeep and ERPClaw handle gift card deferred revenue correctly (the cash hits when the card is sold, the revenue hits when the card is redeemed). ERPClaw includes it in the free version. Bookkeep includes it on every tier.

Is there an even cheaper Bookkeep alternative than ERPClaw?

ERPClaw is $0. There is no cheaper Bookkeep alternative. There is also no “free Bookkeep alternative” with per-transaction posting other than ERPClaw, as far as I know. If you find one, email me and I will update this post.

## Closing

Bookkeep is a cheap, simple Bookkeep alternative to A2X. The reason it is cheap is structural: it posts daily summaries instead of per-transaction detail, and it requires QuickBooks Online or Xero underneath as your real books. That trade is fine for the smallest stores. It stops being fine the moment your CPA wants to trace a refund, your auditor wants to see per-order detail, or you grow past the entry tier.

ERPClaw posts every Shopify order to your books as its own entry, runs on your own machine, replaces QuickBooks entirely, and costs zero dollars a month. Same automation, more depth. Open source. AI-native rather than AI-decorated.

Where to go next:

- Get the Shopify integration: Shopify integration (installable today; App Store listing pending)

- See the install walkthrough: /docs/shopify/install-walkthrough/

- Compare side by side: /compare/bookkeep

- Compare to A2X too: /compare/a2x

- Pricing: /pricing (still $0)

- Source code: github.com/avansaber/erpclaw

Try it for one weekend. If your books do not look better at the end of it, walk away. There is no monthly bill to cancel.

---

## Building an AI-Native ERP With Claude Code
URL: https://www.erpclaw.ai/blog/building-erpclaw-with-claude-code/

# Building an AI-Native ERP With Claude Code
How a spec-first methodology, constitutional financial laws, and Claude Code shipped a full back-office open source ERP a large team would take years to build.

The first artifact of ERPClaw was not code. It was a single large specification, written in plain English. Every table, every action, every naming convention, every validation rule, every test scenario was defined before a single line of Python existed.

That document is the reason an AI-native ERP exists today as a full back-office open source system instead of as a half finished demo. The lesson I want to share in this post is not “Claude Code is amazing” (it is, but that is not the interesting part). The lesson is that AI native is not AI decorated. Decoration is bolting a chat sidebar onto a SaaS product you already shipped. AI native is changing how the software gets built, tested, and priced from the first commit.

If you are a CTO, an engineering manager, or a founder thinking about how to compete with vendors that have a ten year head start, the rest of this post is the playbook I used. It is opinionated, it is reproducible, and it is what made the build economics work.

## My background, and why it matters here

I spent eleven years rolling out SAP and other enterprise systems at Accenture and as an independent architect. I have sat in the rooms where a Fortune 500 retailer paid $50 million for an ERP rollout that ran 18 months late, and I have written the requirements documents that 200 person delivery teams then took 18 months to translate into something resembling working software.

That experience shaped two convictions. First, the bottleneck in enterprise software is almost never coding speed; it is requirements clarity, cross team coordination, and the cost of fixing things that should have been specified up front. Second, mid market companies (the 40 to 500 employee shops) are systematically underserved by the SAPs and Oracles of the world, because the per seat economics do not work below a certain scale.

When Claude Code matured into a tool I could trust for production work in late 2025, both convictions became actionable. The coding bottleneck collapses. The cost of building equivalent scope drops by an order of magnitude. The mid market vacuum is suddenly addressable by a single architect with a good spec and modest infrastructure. That is what ERPClaw is. The point of this post is the methodology, not the product.

## Spec first development, in concrete terms

Most teams treat a specification as a starting point that gets revised heavily during implementation. In spec first development, the specification is the contract. Code is generated from it. When the spec changes, the code is regenerated. When the code drifts from the spec, the code is wrong, not the spec.

For ERPClaw, the spec lives in three layers:

- The master plan. The original specification document covering data model, action catalog, GL semantics, naming conventions, test scenarios, and module boundaries. Nothing in the code exists that is not described here.

- Per module SKILL.md files. Each module has a YAML fronted markdown file under 300 lines that lists every action, its parameters, its return shape, and its tier (basic, intermediate, advanced). This is what Claude Code reads when it generates new actions or fixes existing ones.

- The Constitution. 18 machine readable financial laws (described below) that any module must satisfy, regardless of who or what wrote it.

A SKILL.md entry for a single action looks roughly like this:

`- name: submit-sales-invoice
tier: intermediate
description: Submit a draft sales invoice, posting GL and updating SLE atomically.
args:
invoice_id: { type: string, required: true, format: uuid4 }
posting_date: { type: string, required: false, format: date }
returns:
journal_entry_id: string
gl_balanced: boolean
invariants: [gl_debits_equal_credits, ar_subledger_matches_control]`
That spec block is the source of truth for four audiences at once: Claude Code (which generates the implementation), the test suite (which generates contract tests from the schema), the web dashboard (which renders forms from it), and the human reading the docs. A traditional ERP frontend needs about 150 lines of form code per action. Across the entire action surface that is hundreds of thousands of lines of UI code that simply does not need to exist when the spec drives the surfaces.

The discipline this enforces is what makes the AI assisted coding work. A well specified action produces working code on the first generation about 90 percent of the time. An underspecified action produces plausible looking code that fails on edge cases about 90 percent of the time. The 20 percent of time spent on the spec saves 80 percent of the debugging.

## The Constitution: 18 financial laws, auto validated

The riskiest thing about using AI to build accounting software is that the AI will happily write a general ledger posting function that uses floating point arithmetic. Your trial balance will be off by a penny after a thousand transactions, and nobody will notice until the auditor does.

The fix is not “tell the AI not to do that.” The fix is a constitutional rules engine that rejects any code (human or AI written) that violates a financial law. ERPClaw has 18 of these laws, each expressed as an executable assertion, each enforced at test time across the entire codebase. A short selection:

- Article I: No floats for money. Every monetary column is TEXT, every Python value is a `Decimal`. The validator scans schema definitions and source code; any `REAL` or `float` near a money name fails the build.

- Article III: GL is immutable. The `gl_entry` table has no `updated_at` column. Cancelling a posting creates a mirror reversal entry, never an update. The validator confirms no UPDATE statements target the GL tables.

- Article V: Atomic submission. Every submit action wraps its writes in a single SQLite transaction. The validator parses every submit handler and confirms a `BEGIN ... COMMIT` boundary surrounds the cross table writes.

- Article IX: Twelve step GL validation. Every posting passes through a 12 check pipeline: balanced, no nulls, party set on AR/AP, fiscal year open, account active, currency consistent, and so on. The validator confirms the pipeline is invoked.

- Article XII: Trial balance integrity. After every test run that touches the GL, total debits must equal total credits across the entire database. If the global invariant fails, every GL touching test in the run fails.

If any single article is violated, the offending module cannot ship. There is no human override. This is the regression proof bit: it is impossible to accidentally break double entry bookkeeping and have green tests, because the invariant engine runs over the whole database after every relevant test, not over isolated unit assertions.

This is what makes ERPClaw safe to extend at AI speed. The AI does not need to be perfect. The Constitution will catch it when it is wrong.

## The 80/15/5 theory

Building ERPClaw taught me that ERP module development decomposes into three layers with very different automation profiles:

- 80 percent mechanical. Schema creation, CRUD action scaffolding, naming conventions, audit logging, parameter validation, list and get endpoints. This work is repetitive, well specified, and identical across modules. AI does it perfectly when given the spec.

- 15 percent pattern matching. GL posting patterns, cross module integration glue, report templates, common workflows like draft to submit lifecycle. AI does this well when given the right examples and a checklist of invariants. Human review catches the cross module surprises.

- 5 percent human judgment. Domain edge cases (a partial payment against a multi line invoice with a discount), business specific UX decisions, regulatory subtleties, the questions the AI does not know to ask. This is irreducibly human.

The mistake most teams make with AI coding is treating it as a uniform 100 percent. Either they trust it for everything (and ship floats in money columns) or they distrust it for everything (and waste a decade of free productivity). The right model is to automate the 80, assist the 15, and reserve human attention for the 5 that matters.

This is also what destroys the cost structure of vendors who are not AI native. A 200 person engineering team that spends 80 percent of its hours on mechanical work cannot compete on price with a small team whose mechanical work is automated. The Kodak parallel is exact: it is not that the new product is better, it is that the cost base of the incumbent is no longer defensible.

## What Claude Code does well, and where I override it

After fifteen months of using Claude Code as the primary implementation surface for an open source ERP, here is the honest assessment.

It excels at translating well specified business rules into working code. Give it a SKILL.md entry, the relevant table schema, and the Constitution, and it produces a passing implementation on the first try the vast majority of the time. It does not get bored on action 800. It maintains naming consistency across the entire codebase over weeks of sessions, as long as the spec stays consistent. It is also excellent at test scaffolding, schema migrations, and the unglamorous refactoring tasks (renaming a column across every module, updating a shared library signature) that consume disproportionate human time.

The failure modes matter more than the wins, because they tell you where to spend human attention. Cross module dependencies break first; intercompany invoicing required heavy manual correction because the AI optimised each module locally and missed the global invariants. Edge cases not covered in the spec are a guaranteed regression source: GL reversals with partial payments, garnishment priority ordering, multi currency revaluation when the rate changes mid period. The fix is to add them to the spec the moment you discover them.

Security awareness is approximately zero by default. Claude Code will happily ship your home directory path in an error message or a real Indian taxpayer ID in seed data. I caught 21 such issues in a single audit pass across 220 files; every one was functionally correct and contextually careless. The fix is a security audit of the output, every time, treating AI generated code as if it came from a brilliant but careless junior engineer.

## How I keep the AI honest

Building an open source ERP the size of ERPClaw means generating a lot of code, often unsupervised. The trust model that makes that safe rests on five layers of automated checking, each of which can fail a release independently.

L0 constitutional tests. The 18 articles, plus completeness checks (every Python action documented in SKILL.md), plus structural checks (every module has the required files in the required places).

L2 contract tests. Generated from the SKILL.md specs. Every action is invoked with valid and invalid inputs and the response shape is checked against the schema. This catches drift between spec and implementation immediately.

L3 smoke tests. End to end scenarios that exercise full workflows: quote to cash, procure to pay, hire to retire, manufacturing run with WIP accounting.

Invariant engine (23 checks). Runs after every test that touches financial data. Trial balance balanced, balance sheet equation holds, GL chain hash sequential, no NaN in any financial column, every cancellation has a matching reversal. If any invariant fails, every test in the run fails.

Six gate session pipeline. Local validation, server deploy, vertical install, natural language smoke test, GL integrity, CI status. The session gate is the last line of defence before code reaches a user. The full gate description is on the quality page.

The cumulative effect is a regression prevention checklist that runs in seconds, scales with the codebase, and does not depend on a human remembering to run it. It is one thing to ship a full ERP in a sprint. It is another to keep them shipping correct GL postings six months later, after another wave of new actions has been added.

## The spec is the source of truth, the code is regenerable

The deepest implication of spec first development for an AI native codebase is that the code is no longer the asset. The spec is the asset. The code is a derived artifact.

This sounds esoteric until you watch it play out. When I added a new region (UK PAYE and NI), I did not write a regional payroll module from scratch; I added the regional rules to the spec, regenerated the affected actions, ran the constitutional and contract tests, and shipped. The thinking was already done. When I migrated from one library structure to another, I did not refactor every module by hand; I updated the spec, regenerated the import patterns, and let the test suite tell me what was wrong. A migration that would have been a multi week project on a hand written codebase was a long afternoon.

This regenerability changes the economics. Your competitor’s codebase is the thing they cannot afford to throw away, because they paid millions to write it. Your codebase is something you can rebuild from the spec in a weekend if you find a better architecture.

## What this means for non AI native competitors

A vendor who built a comparable ERP between 2010 and 2025 is now sitting on a code asset with three properties: it cost a great deal to build, it is expensive to maintain, and it is locked into the architectural choices of its era. Their per seat pricing is a function of all three.

An AI native ERP built on the spec first methodology has none of those properties. The build cost was an order of magnitude lower. The maintenance cost is bounded by the spec and the test suite. The architecture can be regenerated when the underlying tools improve.

The implication for pricing is the part most incumbents have not yet absorbed. ERP, CRM, project management, HR, invoicing: these are commodity problems with public business rules. The code was the moat, and the moat was the cost of writing it. That cost is now collapsing. Open source AI native systems will eat commodity SaaS the same way Linux ate proprietary Unix. The vendors who survive will sell domain expertise, regulatory compliance, distribution, and trust. Not code.

## What two weeks of AI assisted building looks like

The original ERPClaw sprint was 14 days, working solo, with Claude Code as the primary implementation surface. Day one was the spec, in full, with no code written. Days two through eleven were two modules per day on average: GL and journals first, then supply chain, operations, payroll and HR, then intelligence and compliance. Days twelve through fourteen were the testing overhaul, the clean install gate, and the security audit that caught 21 findings I should have caught the first time.

The point is not the timeline. The point is the answer to the question every CTO needs to answer: what does your 200 person engineering team do for 18 months that a single architect with a good spec and Claude Code cannot do in two weeks? The answer, mostly, is coordination. AI eliminates the coding bottleneck. Small teams eliminate the coordination bottleneck. Together, that is the order of magnitude.

## FAQ

Is Claude Code production ready for building real software?

Yes, with the right scaffolding. Claude Code on its own is a brilliant junior engineer with no instinct for what should not ship. Wrap it in a constitutional rules engine, a contract test layer, and a session gate, and it becomes a production capable implementation surface. We have built a full production ERP with this setup.

What is the difference between AI native and AI decorated?

AI decorated is bolting a chat sidebar onto a product architected in 2015. AI native is treating AI as the primary implementation layer from day one, which changes the data model (metadata driven), the test strategy (invariant engines), the pricing model (no per seat economics), and the team shape (3 to 5 people, not 200).

Why spec first instead of just prompting harder?

Because prompts do not version, do not test, and do not survive across sessions. A spec is a versioned, testable, reviewable artifact that drives multiple surfaces (AI, API, UI, docs) from one source. Prompt engineering is a tactic; spec first is an architecture.

How is the Constitution different from a linter?

A linter checks syntactic patterns. The Constitution checks semantic invariants across the entire system after the code runs. Article XII (trial balance integrity) cannot be enforced by a linter, because it requires running the test suite and inspecting the resulting database. The Constitution is closer to a property based test framework specialised for financial software.

Can I see the spec and the code?

Yes. ERPClaw is open source and the entire codebase, SKILL.md files, Constitution, and test suite are public on GitHub. The original HackerNoon piece covers the broader story; this post is the engineering view.

## Where to go next

If you are a developer or technical buyer evaluating whether AI assisted coding holds up under the demands of real financial software, the best entry point is the developers page. It walks through the SKILL.md format, the Constitution, and the contract test layer with code samples. The quality page covers the test pyramid and the session gate in detail. The features overview is the product surface; the ERPClaw OS page covers the self extending architecture that grew out of the spec first methodology.

If you want to use the system, `clawhub install erpclaw` is the one command install. The pricing page is short, because the answer is zero for the open source edition. Cloud managed comes later in 2026 for teams that want a hosted instance.

If you are a CTO evaluating what AI changes about the way you ship software, I would rather hear your skepticism than your applause. The methodology in this post is testable. The codebase is open. Reproduce it, break it, or improve it. That is what an AI native open source ERP is for.

---

## ChatGPT for Accounting: Where the Ledger Stops It
URL: https://www.erpclaw.ai/blog/chatgpt-for-accounting-where-the-ledger-stops-it/

# ChatGPT for Accounting: Where the Ledger Stops It
ChatGPT is good at deciding what a transaction means and incapable of being your books. The difference is a system of record, invariants, ownership, and tests.

Short answer. ChatGPT is genuinely useful for the interpretive half of accounting: reading a messy vendor description, proposing a category, explaining a standard, drafting a memo. It cannot be your accounting system, and the reason is not that the model is not smart enough. It is that a chat interface has no system of record, no invariants, no ownership model, and nothing you can unit test. Ask it to post a journal entry and you get text that looks like a journal entry. That distinction is the entire subject of this post. Disclosure up front: we build ERPClaw, an AI-native accounting system, so weigh our framing accordingly.

## What ChatGPT is actually good at here

Start with the honest part, because the case against using a chatbot as your ledger is stronger when it is not padded.

Large language models are good at deciding what something means. “AMZN Mktp US*2K4LM” is a categorisation problem that defeated rules engines for two decades, and a model handles it with context a rule never had. They are good at explaining a treatment you half remember, drafting the narrative around a set of numbers, and turning a policy question into a checklist. If your bookkeeping bottleneck is interpretation, a chat model will help you today.

That is not a small category. Interpretation is the hard, unstructured half of the work, and it is the half that resisted automation the longest.

## The ledger test

Now the other half. Ask a chat model to post a journal entry and look carefully at what comes back.

You get a well-formed journal entry. Debits on the left, credits on the right, plausible accounts, correct-looking amounts. What you do not get is a row in a ledger. Nothing was written anywhere. Nothing was checked against the rest of the book. Nothing prevents the next answer from contradicting this one. You received a very good description of an entry, and the description is the deliverable.

Accounting does not run on descriptions. It runs on a record that persists, balances, and can be audited later by someone who does not trust you. The gap between those two things is not a prompt away.

## Four things a chat model does not have

A system of record. There is no place where a fact lives authoritatively. Two conversations can produce two different answers about the same month and neither is canonical, because neither is a record. Every accounting obligation you have, tax, audit, lender reporting, assumes something canonical exists.

Invariants. This is the one people underestimate, so here is the concrete case. You transfer 100 from savings to checking. A unit test asks: did checking go up by 100? It did. Green tick. But the code forgot the other half, savings never went down, and 100 now exists that did not exist before, with every test passing, because the test only checked what it was told to check.

An invariant asks a different question, after every ledger operation rather than once per test: do all debits still equal all credits across the whole book? That fails immediately, without anyone anticipating this particular bug. In our IEEE IRI 2026 paper we report five bugs that unit tests passed and invariants caught. Each of those five times, the function did exactly what it was asked to do and left the books broken.

Unit tests check that the code did what you asked. Invariants check that the books still make sense. A chat model has neither, and cannot have them, because there is no ledger to check.

An ownership model. In a real system, exactly one component is allowed to write to each table. When a sales invoice is wrong there is one file to open. A chat model has no write path at all, so there is nothing to own, and when the output is wrong the only remedy is to ask again and hope.

Tests. You cannot unit test a prompt. You can evaluate one, which is a statistical claim about a distribution of outputs, and evaluation is genuinely useful. It is not the same guarantee as a test that either passes or fails on a specific input, and an auditor understands that difference even if they would not phrase it that way.

## What actually works: the model decides what you meant, never what the books say

The architecture that resolves this is not complicated to state. The model interprets. Deterministic code writes. Invariants verify.

That split puts the model’s nondeterminism where it is cheap. If it misreads an ambiguous line item, that is visible and correctable. It keeps nondeterminism out of the place where it is expensive, which is the write itself. Every posting goes through code that behaves identically every time, and structural checks run before it lands.

The practical consequence for a buyer: when you evaluate an “AI accounting” product, the question is not how good the model is. It is whether the AI can complete a posting on its own inside a validated transaction, or whether a human still clicks approve on a form. We wrote the full version of that question as a five-trait test for AI-native versus AI-decorated systems, and the shorter framing is in AI-decorated versus AI-native software.

## The model turns out to be the swappable part

One more thing follows from this design, and it surprises people who assume an AI-native system is locked to a model vendor.

Because no module was ever trained, fine-tuned, or prompt-engineered for a specific model, the model is a component rather than a foundation. ERPClaw has run in production through Claude Opus 4.8, Kimi K2.6, Gemini 3.5 Flash, and Codex GPT-5.5, all on the OpenClaw platform, with no change to a single module. If the model were load-bearing for correctness, that swap would be a rewrite. It is a configuration change instead, precisely because correctness lives in the deterministic layer and the invariants.

The inverse is worth stating too: a product whose accuracy depends on which model is behind it has told you where its guarantees come from.

## So should you use ChatGPT for accounting?

Use it for what it is good at. Ask it to explain a revenue recognition question, to draft the note that accompanies a variance, to propose categories for a pile of unfamiliar transactions. That is real leverage and it costs nothing to try.

Do not ask it to be the books. Not because it is unhelpful, but because the thing you need from an accounting system is the part a chat interface structurally does not provide: a record that persists, checks that run whether or not anyone thought to write them, and a single place responsible for every number.

If you want that with the interpretive half included, that is what an AI-native system is for. Ours is free, open source, and self-hosted, so you can read the invariant engine before you believe any of this: see AI accounting for the capability picture and AI bookkeeping for the day-to-day mechanics.

Tagschatgptai-accountingbookkeepingarchitecturellm
Share:XLinkedInHacker News

---

## The Clearing Account Pattern, in Plain English
URL: https://www.erpclaw.ai/blog/clearing-account-pattern-shopify-stripe/

# The Clearing Account Pattern, in Plain English
The clearing account pattern is the trick that makes Shopify and Stripe deposits actually balance to your sales. Here is how it works, in plain English.

You sold $5,200 of stuff on Shopify last week. Shopify deposited $4,318.62 in your bank. Those two numbers will never be equal, no matter how long you stare at the screen, and that is not a bug. The fix is one of the oldest patterns in accounting and it has a name almost nobody outside the bookkeeping world has heard of: the clearing account pattern.

This post is the plain-English explainer. We will cover what a clearing account actually is, why Shopify and Stripe both force you to use one whether you realize it or not, walk a single $100 sale through the entire money flow, and show what breaks when the pattern is missing. By the end you will know more about the clearing account pattern than 95% of Shopify and Stripe merchants, and you will know how to spot when yours is broken.

## The thing nobody tells you about online payments

The mental model most founders start with is simple. A customer pays you $100. The $100 lands in your bank. You owe tax on $100. Easy.

Reality is messier. When a customer pays $100 on your Shopify or Stripe checkout, the money does not go straight to your bank. It goes to Shopify or Stripe first. They hold it for a couple of days. They take their cut. They net out any refunds, chargebacks, or app fees. Then they batch up several days of activity into one deposit, called a payout, and wire that to your bank.

So at any given moment, there is real money that legally belongs to you but is sitting in someone else’s account. That money is not in your bank yet. It is also not “gone.” It is in transit. You need a place on your books to track it. That place is a clearing account.

Without one, your books always look broken. With one, they balance to the cent.

## What a clearing account actually is

Forget the accounting textbook for a second. A clearing account is a holding bucket on your books. It sits between Shopify or Stripe and your real bank account. Money flows into the bucket when a sale happens. Money flows out of the bucket when the deposit hits your bank.

Picture a wooden box on a shelf in your office labeled “Money Shopify owes me.” Every time someone buys a t-shirt, you write a slip and drop it in the box. Every time Shopify pays you, you take the matching slips out and clip them to the bank deposit. If the box is empty at the end of the month, your books match Shopify’s. If there are slips left over, that is money still owed to you. If the box goes negative (you took out more than you put in), something was double-counted.

The wooden box is the clearing account. That is the entire idea. The accounting textbook calls it a “transit account” or a “suspense account” or sometimes an “undeposited funds” account. The name does not matter. The pattern matters.

The reason this pattern is so widely used in real bookkeeping is that it gives you a built-in lie detector. Every transaction has to come into the bucket and leave the bucket. If the bucket does not zero out, you know something is wrong before your accountant or the tax authority does.

## The canonical $100 sale, drawn out

Let us walk a single $100 Shopify sale through the bucket, step by step, with the math out loud. The same logic applies to Stripe, just with different deposit timing.

### Step 1: The customer hits Pay

Customer buys a $100 t-shirt. Shopify Payments will keep about $3 in transaction fees and owe you the remaining $97. The order shows up in your Shopify admin instantly.

On your books:

- Sales for the day go up by $100.

- The “Shopify owes me” bucket goes up by $97.

- Shopify fees expense goes up by $3.

The bucket now holds $97. Notice that nothing has hit your bank yet.

### Step 2: Shopify holds the money for a couple of days

Standard US payouts settle 2 to 3 business days after the sale. During those two days, nothing changes on your books. The $97 just sits in the bucket. Your bank balance is still where it was. That is normal.

### Step 3: Shopify wires the payout

Three business days later, $97 lands in your real bank account.

On your books:

- Bank account goes up by $97.

- The “Shopify owes me” bucket goes down by $97.

The bucket is now empty. Fully reconciled. The $100 sale, the $3 fee, and the $97 deposit all line up perfectly. You can prove every dollar to the IRS, your accountant, your investor, or anyone else who asks.

### Now scale it to a real store

A real Shopify store does this hundreds of times per day, mixed with refunds, chargebacks, gift cards, and capital repayments. The exact same pattern handles all of it. The bucket fills up during the day, drains down each time a payout settles, and should always trend toward zero. The only money that should be sitting in the bucket at any given time is the last 2 to 3 days of sales that have not been paid out yet. That is called in-flight money, and it is normal.

If you want the formal version with debits and credits, the Shopify integration architecture page walks through the actual journal entries our connector posts. For this post, the bucket metaphor is the whole story.

## Why Stripe needs the same pattern

Everything above applies to Stripe too. The numbers and the timing change, but the logic is identical.

A Stripe sale moves through the same three stages. The customer pays $100. Stripe takes its 2.9% plus 30 cents (so about $96.81 to you). Stripe holds the money in your Stripe balance for the standard 2-day rolling window for new accounts (or 7 days, or T+0, depending on your country and account type). Then Stripe batches up several charges into a payout and wires it to your bank.

If you have a Stripe Clearing account on your books, every successful charge fills the bucket and every payout drains it. Subscription renewals do the same thing. Refunds reverse the original entry in the bucket. Disputes hold money in reserve until they resolve. Application fees on Stripe Connect get netted out the same way Shopify Capital does.

The reason this matters even more for Stripe than for Shopify is that Stripe is often used for SaaS, marketplaces, and services where the gross-revenue number matters for ASC 606 revenue recognition, MRR reporting, and investor metrics. If you book the bank deposit as your revenue, your MRR is wrong, your gross margin is wrong, and your audit will be painful. The Stripe clearing account is what keeps gross revenue and net deposits clearly separated.

The Stripe integration architecture page goes deeper on the GL postings if you want to see the actual entries.

## A2X invented this for ecommerce, but the pattern is older than the internet

Quick history note. The clearing account pattern has been in accounting textbooks for at least 80 years, used by every business that ever waited for a check to clear. What was new, around 2014, was applying the pattern in software, automatically, to Shopify payouts.

The credit goes to A2X. They were the first tool to look at the mess of “Shopify deposit minus fees minus refunds minus reserves” and say: stop trying to match the deposit to your sales. Match the deposit to the bucket, and match the bucket to your sales separately. Their summary-journal-entry approach became the de facto standard for serious ecommerce accounting, and the same approach is now used by Synder, Bookkeep, Webgility, and basically every payments accounting tool that ships in 2026. If you are weighing the two biggest names against each other, A2X vs Synder compares them head to head.

## What breaks the clearing balance

The clearing account is your truth-teller. When it does not zero out, something is wrong. After helping a lot of stores debug this, the broken clearing balance almost always comes from one of four causes.

### 1. Manual bank entries that bypass the bucket

Someone records the Shopify deposit straight to “Sales” instead of routing it through the clearing account. Now the bucket has the original sales sitting in it forever, and the bank deposit is double-counted as revenue. Classic mistake when a bookkeeper joins mid-year and was not told the pattern.

The fix: delete the manual bank entries and let the Shopify connector handle the deposit. If you really need to enter it by hand, post it as a transfer from the clearing account to the bank, not as new income.

### 2. Refund duplication

A customer returns a $50 item. Shopify refunds them. Your accounting tool catches the refund webhook and reverses the original entry, then your bookkeeper sees the negative number on the bank statement and reverses it again. The clearing account is now $50 short.

The fix: pick one source of truth for refunds and trust it. Do not record refunds twice.

### 3. Dispute reserves not handled

A customer disputes a $200 charge. Stripe or Shopify pulls $200 plus a $15 dispute fee out of your next payout and parks it as a reserve until the dispute resolves. If your books only know about “sales” and “fees,” that $215 looks like a missing payout.

The fix: have a separate Dispute Reserve account on your books. When the dispute resolves in your favor, the reserve releases back into the clearing account. When you lose, the reserve becomes an expense.

### 4. Currency conversion

You sold in EUR. Stripe paid you in USD. The conversion rate Stripe used and the rate your bank used disagree by a fraction of a percent. The clearing account is now $1.71 off. This will happen forever and is not an error. Record the difference as a foreign exchange gain or loss and move on.

## How to fix a broken clearing balance

If your Shopify Clearing or Stripe Clearing account is sitting at a non-zero balance and it is not just in-flight money, here is the order to debug.

- Pull the clearing account ledger for the last 30 days. Look at the running balance day by day.

- Find the day it stopped zeroing out. That is your culprit window.

- Compare every entry in the bucket against the source. Pull Shopify or Stripe’s reports for the same window. Every entry in the bucket should map to a real event on the platform side.

- Look for the four causes above. Manual bank entry? Refund booked twice? Dispute reserve missing? Currency conversion?

- Post a correcting entry, not a fudge. Resist the urge to “plug” the balance with a journal entry to “Other Income.” That hides the problem and will repeat next month. Find the root cause, post the correction with a note explaining why, and move on.

For most stores, the first time you do this exercise it takes an afternoon. After that, if you have automation in place, the clearing account stays clean by itself.

## Automating the pattern

The pattern is universal, but doing it by hand at any real volume is painful. ERPClaw runs the clearing account pattern automatically for both Shopify and Stripe, on hardware you control. You pair it to your store and your Stripe account, and from then on every order, payout, refund, dispute, and reserve flows through a clearing account you can audit at the per-transaction level. Because ERPClaw is AI-native, you ask the assistant “show me the Stripe clearing balance for last week and break out anything older than 3 days” in plain English instead of clicking through a reports menu. The Shopify features page and Stripe features page cover the install; the pricing page is mostly empty.

Most merchants check the clearing balance once a week, see “$0.00 balanced” or “$1,247 in-flight (last 2 days),” and move on. That is the whole reconciliation. No spreadsheet, no Saturday morning, no surprises in April.

## FAQ

### What is a clearing account in plain English?

A clearing account is a holding bucket on your books that sits between a payment processor (like Shopify or Stripe) and your real bank account. Sales fill it up. Payouts drain it down. When the bucket is empty, you are fully reconciled. The pattern is used so the deposit on your bank statement does not have to match your sales total directly.

### Is a Shopify clearing account the same as a Stripe clearing account?

The pattern is identical. The accounts are separate (you would normally have one Shopify Clearing account and one Stripe Clearing account, so each platform’s activity stays auditable on its own). Both follow the same logic: sales in, fees and refunds out, payouts out, balance trends to zero.

### Where does the clearing account live in QuickBooks or Xero?

You create it as a current asset account, usually called “Shopify Clearing” or “Shopify Payments Clearing” (and the same for Stripe). It is treated as money owed to you, so it sits next to your bank accounts on the balance sheet. The official Shopify Payments documentation explains the payout structure that the clearing account is designed to absorb.

### Why does my clearing account never quite zero out?

Two reasons, one normal, one not. The normal one is in-flight money: sales from the last 2 to 3 days that have not been paid out yet. That should always be there during business hours. The not-normal one is one of the four causes above (manual bank entry, refund duplication, dispute reserve, currency conversion). If the balance is bigger than 2 to 3 days of typical sales, you have a real error to chase.

### Do I need a clearing account if I am a tiny store?

Honestly, yes. Even at 10 orders a month, the moment you take your first refund or your first chargeback, the math without a clearing account stops working. Set up the pattern from day one and it will scale with you to 10,000 orders.

### Can I do this with just spreadsheets?

You can, and a lot of solo founders do at the start. The pattern is the same. You just maintain the bucket in a sheet column and tick off entries as payouts settle. It works fine up to about 50 orders a month. Past that, you want either middleware (A2X, Synder, Bookkeep) or an integrated system (ERPClaw) doing it for you. The official Stripe Payouts documentation is worth reading regardless, since it explains exactly when money moves and why.

### Is the clearing account different from “undeposited funds” in QuickBooks?

Same idea, different name. QuickBooks calls a generic clearing account “Undeposited Funds” by default. Shopify- and Stripe-specific clearing accounts are usually created as separate current asset accounts so each platform’s activity stays separated. That separation is what lets you audit each platform on its own.

## Closing thought

The clearing account pattern is one of those ideas that sounds boring until you have lived through the alternative. Once you have spent one Saturday morning trying to make a Shopify deposit equal a sales total that it was never going to equal, the bucket metaphor clicks and never leaves you. It is the pattern that turns “my books always look broken” into “my books balance to the cent every week.”

If you want it running automatically against your Shopify store or Stripe account, the Shopify integration and Stripe integration ship with the pattern built in. If you would rather keep using your existing tools, just set up the clearing account in QuickBooks or Xero and route every Shopify or Stripe entry through it. Either way, the bucket is the answer.

---

## Shopify OAuth on Cloudflare Workers: Our Pairing Bridge
URL: https://www.erpclaw.ai/blog/cloudflare-worker-shopify-oauth-pairing/

# Shopify OAuth on Cloudflare Workers: Our Pairing Bridge
How we built a Cloudflare Worker bridge between Shopify's embedded App Bridge OAuth and a self-hosted CLI ERP, with the gotchas we hit along the way.

A Shopify app has to live inside the Shopify admin. A self-hosted ERP lives on the customer’s own machine, behind their own firewall, with no public URL. Those two architectural requirements collide head-on, and how you resolve them is the difference between a five minute install and a multi-week support headache.

This is the engineering write-up of how we built the ERPClaw Shopify pairing bridge on Cloudflare Workers. Token Exchange, HKDF-derived per-shop secrets, the App Bridge gotchas, the things the Shopify docs do not tell you, and why we ended up here after rejecting three other approaches.

This post is for developers building Shopify apps that need to bridge to off-platform infrastructure. If you are also wrestling with “the embedded admin requires Token Exchange but I do not want to host a backend,” this is what we did.

## The architectural problem in one paragraph

Shopify embedded apps run inside an iframe in the Shopify admin. They authenticate via Shopify App Bridge using session tokens, and they exchange those session tokens for offline access tokens via Token Exchange. The offline token is what you use to call the Admin API for that shop.

ERPClaw is a self-hosted ERP. It lives on the customer’s machine. There is no public URL we can point Shopify at. We cannot host the offline tokens because each customer’s data should stay on their machine. So how does the customer’s local ERPClaw process get the offline token without us, the app developer, hosting any per-customer infrastructure?

The answer is a stateless Cloudflare Worker that handles the OAuth dance, derives per-shop encryption keys deterministically, and hands the token to the customer’s CLI via a one-time pairing code. We never store the token. We never see the customer’s data. The Worker is a pure broker.

## Why we rejected the obvious approaches

Before settling on Workers, we ran through three other approaches and rejected each. Useful for context.

Option 1: host a multi-tenant SaaS backend. Stand up a normal app server, store offline tokens per shop, serve as the Shopify app’s backend. Standard pattern. The reason we rejected it: it makes ERPClaw not really self-hosted anymore. The whole product positioning is “your data on your machine, not our cloud.” Hosting offline tokens centrally turns us into a SaaS company in everything but name. Also costs money to run. Also makes us responsible for breach risk on every customer’s Shopify data.

Option 2: have customers run a publicly accessible web server. Tell customers to expose their ERPClaw install on a public URL with TLS, and have Shopify call that directly. The reason we rejected it: small business customers do not run public web servers. They run a Mac on their desk. Asking them to set up port forwarding, a domain, a TLS certificate, and a static IP is a non-starter. Even with ngrok or Cloudflare Tunnel, the operational burden is too high for the install-in-five-minutes promise.

Option 3: use Shopify’s offline token in the embedded app and never sync to the local ERP. Just build the entire Shopify app as a web app and skip the local sync. The reason we rejected it: the whole point of the integration is that your accounting data lives in your local ERPClaw install, not in Shopify. The Shopify side is a thin source of truth for orders and customers; the local ERPClaw is where the GL, the chart of accounts, the AR aging, and the financial reports live.

So we needed a fourth option. A stateless broker that handles the OAuth dance, hands the token to the local CLI, and never holds long-lived state. That is the Cloudflare Worker.

## The architecture

Three components:

- Cloudflare Worker (`erpclaw-shopify-bridge`) at a fixed URL. Stateless. Handles the OAuth dance with Shopify, derives per-shop encryption keys via HKDF, and brokers the pairing handoff.

- Shopify embedded app (Cloudflare Pages) that loads inside the admin iframe. Calls App Bridge, runs Token Exchange, displays the pairing code to the merchant.

- Local ERPClaw CLI on the customer’s machine. Calls a `pair-shopify-shop` action with the pairing code. The CLI talks to the Worker, gets the offline token, stores it locally encrypted, and never talks to the Worker again for that shop.

The flow:

`Merchant clicks "Install" in Shopify App Store
↓
Shopify redirects to embedded app URL
↓
Embedded app loads in iframe
↓
App Bridge produces session token
↓
Embedded app calls Worker /pair endpoint with session token
↓
Worker calls Shopify Token Exchange API with session token
↓
Worker gets offline access token from Shopify
↓
Worker derives a per-shop encryption key via HKDF(master_secret, shop_domain)
↓
Worker generates a 6-character pairing code, encrypts the offline token with the per-shop key
↓
Worker stores the encrypted token in KV with the pairing code as the key, TTL 10 minutes
↓
Worker returns the pairing code to the embedded app, which shows it to the merchant
↓
Merchant runs `erpclaw shopify pair --code XYZ123` on their local machine
↓
CLI calls Worker /redeem endpoint with the pairing code
↓
Worker derives the per-shop key (same input, same output, no storage needed)
↓
Worker decrypts the offline token, returns it, deletes the KV entry
↓
CLI stores the offline token locally encrypted by the merchant's machine identity
↓
CLI confirms pairing complete to Worker, syncs first batch of orders`
After this flow, the Worker never sees this merchant’s data again. The offline token lives only on the customer’s machine. The KV entry is gone. The pairing code is single-use.

## Why HKDF for per-shop keys

If we encrypted every offline token with the same global key, a breach of that key would compromise every shop’s data. If we generated a random per-shop key, we would have to store it somewhere, which puts us back to “we hold per-customer state.”

The compromise is HKDF. We hold one master secret. The per-shop encryption key is `HKDF(master_secret, shop_domain)`. The same shop always derives the same key. We never store the per-shop key; we re-derive it on demand. A breach of the master secret compromises everything (which is true of any encryption scheme), but the keys themselves are derived deterministically without storage.

In Worker code:

`async function deriveShopKey(masterSecret: ArrayBuffer, shopDomain: string): Promise<CryptoKey> {
const salt = new TextEncoder().encode("erpclaw-shopify-v1");
const info = new TextEncoder().encode(shopDomain);

const baseKey = await crypto.subtle.importKey(
"raw", masterSecret, "HKDF", false, ["deriveKey"]
);

return crypto.subtle.deriveKey(
{ name: "HKDF", salt, info, hash: "SHA-256" },
baseKey,
{ name: "AES-GCM", length: 256 },
false,
["encrypt", "decrypt"]
);
}`
The salt is constant across shops; the info parameter is the shop domain, which is what makes the derived key shop-specific. This is the standard HKDF pattern from RFC 5869.

The master secret lives in Cloudflare Workers Secrets, encrypted at rest, never exposed in logs or environment dumps. Rotation is possible via a versioned salt (`erpclaw-shopify-v2`) which would invalidate all existing pairings and force re-pair, an acceptable operational tradeoff for an emergency rotation.

## Token Exchange, the actual call

Token Exchange is Shopify’s OAuth 2.0 token exchange flow per RFC 8693. The embedded app gets a session token from App Bridge (a JWT signed by Shopify), and the Worker exchanges it for an offline access token by POSTing to the shop’s token endpoint.

`async function exchangeForOfflineToken(
shopDomain: string,
sessionToken: string,
apiKey: string,
apiSecret: string
): Promise<{ access_token: string; scope: string }> {
const response = await fetch(
`https://${shopDomain}/admin/oauth/access_token`,
{
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({
client_id: apiKey,
client_secret: apiSecret,
grant_type: "urn:ietf:params:oauth:grant-type:token-exchange",
subject_token: sessionToken,
subject_token_type: "urn:ietf:params:oauth:token-type:id_token",
requested_token_type: "urn:shopify:params:oauth:token-type:offline-access-token",
}),
}
);

if (!response.ok) {
throw new Error(`Token exchange failed: ${response.status}`);
}

return response.json();
}`
The catch is that the session token must be validated before calling Token Exchange. If a malicious merchant page submits a fake session token, the Worker should reject it before contacting Shopify. Validation means verifying the JWT signature against the Shopify-provided JWK set, checking the `aud` claim equals our app’s API key, checking the `iss` claim matches the shop domain, and checking the `exp` claim is in the future.

`async function validateSessionToken(
sessionToken: string,
apiSecret: string,
expectedShop: string
): Promise<JWTPayload> {
const payload = await jwtVerify(sessionToken, apiSecret, {
audience: API_KEY,
issuer: `https://${expectedShop}/admin`,
});

if (payload.exp * 1000 < Date.now()) {
throw new Error("Session token expired");
}

return payload;
}`
We learned this the hard way during the 2026-04-25 debug session. Without strict validation, the Worker happily forwards any session token to Shopify and Shopify happily issues an offline token if the JWT decodes. That is a bad day waiting to happen.

## The cached-token gotcha

The single hardest bug in the pairing flow was a cached token issue that took most of a day to track down. Symptom: the merchant clicks Install, the embedded app loads, the Worker returns a pairing code. The merchant runs the CLI command. Nothing happens. Re-paired three times. Same result.

Root cause: App Bridge was returning a cached session token from the previous install attempt. The Worker validated the cached token (still valid, not expired), called Token Exchange, got a stale offline token from Shopify (which Shopify legitimately invalidates after Install Re-runs), and stored that in KV. The CLI redeemed it, got a 401 from Shopify on the first API call, and retried, and retried, and retried until it gave up.

The fix was twofold:

- Force fresh session token on every embed load. App Bridge has a `forceRefresh: true` option on `getSessionToken()` that bypasses the cache. We always pass it on the pairing flow.

- Reject session tokens older than 30 seconds at the Worker. Even if App Bridge returns a “fresh” token that is actually 10 minutes old, the Worker checks the `iat` claim and refuses anything older than 30 seconds. Belt and suspenders.

This is the kind of thing the Shopify docs do not call out. The session token is a JWT and JWTs have an expiration claim, but App Bridge silently caches them and there is no mention of cache invalidation in the embedded app docs. We learned by reading the App Bridge source on GitHub.

## Cloudflare Pages for the embedded app

The embedded app itself (the part that loads in the iframe) is a static site on Cloudflare Pages. Astro build, no server-side rendering, just HTML and JS. App Bridge, Token Exchange call to the Worker, display pairing code, done.

Static site means we have no per-shop server-side state in the embedded app either. The whole architecture is stateless except for the 10-minute KV entry holding the encrypted pairing code.

Things we hit on Pages:

- Custom domain on Pages requires the Cloudflare Pages app to own the zone. If the zone is on a different account, the routing breaks silently. We unified the zone and the Pages project under one account.

- App Bridge requires HTTPS and a fixed redirect URL. Pages provides both for free.

- The Shopify app config in Partners dashboard must list the exact embedded app URL. Mismatches produce a generic “App URL invalid” error with no detail. Triple check the URL.

## Why we put this on Cloudflare instead of AWS Lambda

Three reasons.

Latency. Cloudflare Workers run at the edge globally. Shopify shops are distributed worldwide. A merchant in Australia hitting an AWS us-west-2 Lambda gets 200ms+ added to every request. A Cloudflare Worker is ~30ms from anywhere.

Cost. Workers have a generous free tier (100k requests/day) and the pairing flow is a few requests per merchant per install. We are paying $0 for the bridge today and will be paying $0 at 10,000 merchants.

KV TTL. Cloudflare KV has TTL built in. The 10-minute pairing code expiry is `expirationTtl: 600` on the put. No cron job needed to clean up expired codes. AWS DynamoDB has a TTL feature too but it is approximate (up to 48 hours late) and not appropriate for “the pairing code MUST be unusable after 10 minutes.”

The downside of Workers is the runtime constraints. No Node-style filesystem, limited libraries, ESM only, 50ms CPU time limit per request. For a pairing broker this is fine. For the actual ERPClaw app, this would not work, which is why the ERP itself runs locally and not on Workers.

## What this saves the customer

From the merchant’s perspective, the install flow is (the integration is shipped and installable today, with the public App Store listing pending; see the feature tour or the setup docs):

- Click “Install” on the listing. (Shopify side, ~15 seconds.)

- App loads in admin, shows a 6 character pairing code. (~3 seconds.)

- Run `clawhub install erpclaw` on local machine if not already installed. (~30 seconds.)

- Run `erpclaw shopify pair --pairing-code XYZ123`. (~5 seconds.)

- Done. ERPClaw is now syncing orders.

Total install time: under two minutes. No webhook configuration. No API key handling. No public URL setup. The merchant never touches a credential.

The Worker did the work the merchant did not have to do.

## Open source the bridge?

The Worker code is in a private repo today. We may open source it once the Shopify app is publicly listed. The cryptographic primitives (HKDF, AES-GCM via WebCrypto, Token Exchange per RFC 8693) are all standard. The only secret is the master secret in Workers Secrets, which we rotate independently.

If you are building a similar bridge for a different platform (Stripe Apps, Square, Xero), the pattern is reusable. Stateless Worker, HKDF for per-tenant keys, short-lived KV for pairing codes, push the long-lived state to the customer’s machine.

## Try the integration

The Shopify integration is shipped and installable today; the feature tour shows what it does. The Stripe integration is live on the Stripe Marketplace. Both use the same architectural pattern (broker for the OAuth dance, customer holds the data) for the same reason.

- Read the Shopify integration docs

- Install ERPClaw

- Read the Shopify integration announcement

- Read about the engineering architecture

- SQLite or PostgreSQL: the database choice

ERPClaw is open source, runs on SQLite or PostgreSQL via PyPika, and the Stripe + Shopify integrations are deep, audit-grade, and free.

## FAQ

### Can I reuse this Worker bridge pattern for my own Shopify app?

Yes, the pattern is reusable. The components are: a Cloudflare Worker (or any edge function) that handles Token Exchange, an HKDF derivation for per-shop keys, a short-lived KV/Redis entry for the pairing code, and a CLI on the customer’s machine that redeems the code. None of the pieces are ERPClaw-specific.

### What happens if the merchant loses the pairing code?

The code expires in 10 minutes regardless. The merchant goes back to the Shopify admin, reloads the app, and gets a new pairing code. Token Exchange re-runs and a fresh offline token is issued by Shopify. No state to clean up.

### How do you handle Shopify webhook delivery if the customer’s machine is behind a firewall?

Webhooks are delivered to the Worker, not directly to the customer’s machine. The Worker writes the webhook to a queue (Cloudflare Queues or KV with a known prefix) and the customer’s local CLI polls the Worker on a schedule (every 30 seconds by default) to pull pending webhooks. This is the inverse of the typical webhook flow but works without exposing the customer’s machine.

### Is the offline token at risk if the Worker is breached?

The offline token is encrypted in KV with the per-shop derived key. The master secret is in Cloudflare Workers Secrets, separate from the Worker code. A breach of the Worker code alone does not expose tokens; an attacker needs both the code and the master secret. We treat the master secret as the only crown-jewel credential and rotate via versioned salt if needed.

### Why not use Shopify’s official Node template?

The official template assumes you are hosting a multi-tenant backend that holds offline tokens centrally. We are not. The architecture is fundamentally different because ERPClaw is self-hosted, not SaaS.

### How does ERPClaw handle Shopify rate limits?

The Shopify API has a rate limit of 2 requests per second per shop (40 in burst). The local ERPClaw CLI implements exponential backoff plus a token bucket per shop. The Worker does not call Shopify on the merchant’s behalf except during pairing, so it does not hit shop-specific rate limits.

### Is the bridge open source?

Not yet. The cryptographic logic is standard (HKDF, AES-GCM, Token Exchange per RFC 8693) and the architecture is described above. We may open source the Worker code once the public App Store listing is live.

Install ERPClaw and try the Stripe and Shopify integrations today; the Shopify docs cover setup.

---

## Shopify Cost of Goods Sold: Your Real Margin Number
URL: https://www.erpclaw.ai/blog/cost-of-goods-sold-shopify-store/

# Shopify Cost of Goods Sold: Your Real Margin Number
Most Shopify stores miscalculate cost of goods sold by 10-20%. Here is what actually counts as COGS, what Shopify misses, and how to fix it for free.

You think your gross margin is 65%. Your bank account thinks it is 48%. One of you is wrong, and it is not the bank account.

This is the most common spreadsheet lie in Shopify land, and it almost always traces back to a wrong number for Shopify cost of goods sold. For example, you priced a t-shirt at $30, your supplier charges $10, so you wrote down “67% margin” and went back to scaling ads. But the shirt also costs $2.40 in inbound freight, $1.20 in customs duty, $0.90 in Shopify Payments fees, and roughly $1.80 once you average in returns. Real cost is closer to $16.30. Real margin is 46%, not 67%.

Multiply that gap across 10,000 orders and you have built a business on the wrong number. This post covers what should go into your Shopify cost of goods sold calculation, what Shopify leaves out, why the gap matters at loan time, and how to fix it without a bookkeeper.

## What Counts as Cost of Goods Sold for a Shopify Store

Cost of goods sold is the money you actually spent to put the product into the customer’s hands. Not your office rent. Not your Klaviyo bill. Not your founder salary. The cost that is directly tied to the unit you just sold.

For a Shopify store selling physical product, the honest list is shorter than people think and longer than Shopify’s “Cost per item” field suggests:

Always counts as COGS:

- The wholesale price you paid your supplier, manufacturer, or print-on-demand partner

- Inbound freight to get the product into your warehouse (sea freight, trucking, the FedEx bill on the case of mugs from Ohio)

- Import duty, customs broker fees, and tariffs

- Per-unit packaging that is part of the product (the candle’s tin, the shirt’s hangtag, the gift box if it ships in every order)

- Direct labor that touches the product, if you have it (the seamstress who hems the shirts, not your VA)

Sometimes counts, depends on how you treat it:

- Outbound shipping to the customer. If you charge for it, most accountants put it in COGS. If it’s free shipping baked into the price, it has to come out of margin somewhere, and COGS is the cleanest place

- Payment processing fees (Shopify Payments, PayPal, Shop Pay Installments). Some put these in operating expenses, but for honest unit economics, they belong in COGS

- Returns and refunds. The product comes back, sometimes resellable, often not. The unrecoverable portion is a COGS adjustment

Never counts as COGS:

- Marketing, ads, influencer fees

- Software subscriptions (Shopify itself, Klaviyo, Recharge, etc.)

- Office rent, utilities, founder salary

- Accountant fees, legal fees, your CPA’s bill

The IRS Schedule C definition of COGS is a useful gut-check here, and the IRS publishes a clear breakdown of what they expect to see. If a cost would not exist if you had sold zero units, it is probably not COGS.

## What Shopify’s “Cost per Item” Field Actually Tracks

Shopify gives you one field per product variant called Cost per item. You enter the wholesale price you paid, and Shopify uses it to calculate a profit number in your reports. For example, if you sell a $30 shirt and you entered $10 in Cost per item, Shopify reports $20 of profit per unit and a 67% margin.

That number is wrong. Or more accurately, it is incomplete in a way that always flatters you.

Shopify’s own help docs confirm the field only captures the unit cost you type in. It does not include:

- Inbound freight. For example, if you import 1,000 mugs and the shipping was $1,800, that is $1.80 per mug, every mug, forever. Shopify will not add it.

- Customs and duties. Same story. A 10% tariff on imported apparel is real money. Shopify ignores it.

- Multi-warehouse cost differences. You buy mugs from two suppliers. One charges $4, one charges $5.20. Shopify shows one Cost per item, so half your COGS reports are off by 30%.

- FIFO vs average vs the last invoice. Your supplier raised prices three times this year. Shopify still shows whatever you typed in last. The mugs you actually sold today may have cost you $3.80 (old batch) or $5.20 (new batch). Shopify cannot tell.

- Returns. A returned shirt is COGS that should reverse. Shopify does not adjust the cost figure when a refund comes back.

So the “67% margin” you see in Shopify Analytics is a starting point, not the answer. Shopify COGS tracking is a feature Shopify gives you the data hooks for and expects your accounting tool to finish the job.

## FIFO vs Average Cost, in Plain English

You do not need an accounting degree to run a store, but you do need to know which mug got sold so the right cost lands against it.

FIFO is “first batch in, first batch out.” Say you bought 500 mugs at $4 in January and 500 at $5.20 in March. When a customer buys one in April, FIFO says you sold a January mug and COGS for that order is $4.

Weighted average blends the two batches. Total cost was $4,600 for 1,000 mugs, so every mug is treated as $4.60 regardless of when it shipped.

Why does it matter? Because the IRS cares, and because your gross margin numbers can swing 5-15% during a price-rising period depending on which method you pick. For most clothing, candle, and mug brands, FIFO matches what actually happened on the shelf.

The deeper trade-off is in our companion post on FIFO vs weighted average. Short version: pick one, stick with it, and make sure your tools actually do the math. Shopify alone does not.

## Why Wrong COGS Numbers Will Bite You

You can run a store on a wrong COGS number for a long time. The day it matters, it really matters.

SBA loan or line of credit. Banks underwriting a working-capital loan want to see a real gross margin number on your last two years of tax returns. If you reported COGS based on the Shopify field and the bank’s analyst recalculates including freight and duties, your margin shrinks, your debt-service-coverage ratio drops, and your loan offer gets significantly cut or denied.

Investor due diligence. Same problem, sharper teeth. A buyer or VC running diligence will rebuild your unit economics from invoices. If your reported margin and your reconstructed margin disagree, you lose trust and you lose significant valuation.

Your own pricing decisions. This is the quiet one. You think a SKU is your best margin product so you push it in ads, but it is actually a freight-heavy import you have been losing money on. You discount a “high-margin” item in a Black Friday sale and find out in January it was your worst performer all along. Shopify gross margin numbers fed by partial COGS data drive the wrong calls every day.

Tax over- or under-payment. Understate COGS and you overpay tax on phantom profit. Overstate it and you eventually get audited. Either way, COGS is one of the IRS’s top audit triggers for ecommerce.

## Why Multi-Warehouse Stores Get Wrecked the Hardest

If you sell from one warehouse, the cost story is messy but workable. If you ship from two or three (a 3PL on each coast, your own basement plus an Amazon FBA pool, print-on-demand alongside your own stock), the math turns into a mess no spreadsheet keeps up with.

Same SKU, two warehouses, two prices: which one did this order ship from, and what did that specific unit cost? You need to calculate COGS Shopify at the location level, not the global level. Multi-warehouse COGS is one of the things Shopify’s built-in reporting does not even attempt; it shows you one cost per variant and calls it a day.

This is exactly the gap most paid Shopify accounting apps charge premium-tier money to fill. A2X, for example, gates COGS tracking to its higher tier and still does not break it out per warehouse on the standard plans (see our full breakdown on the A2X comparison page). For a brand running $1M+ through two locations, that gap is real money.

## How to Fix Your COGS Without Hiring a Bookkeeper

You have three options for getting an honest COGS number on a Shopify store.

Option 1: Spreadsheet it manually. Export your orders, export your inventory cost adjustments, calculate freight allocation per unit, recalculate at the end of every month. This works at very small scale (under 200 orders/month), eats one full day per month, and breaks the moment you add a second warehouse.

Option 2: Pay a bookkeeper plus a paid Shopify accounting app. A2X, Bookkeep, Synder, and similar tools sit between Shopify and QuickBooks or Xero. The best of them do per-transaction journal entries, payout reconciliation, and gift card deferred revenue. The catch: typically $29 to $229/month for the app, plus your bookkeeper’s hours, plus you still need premium tiers for COGS. You are looking at $400-1,500/month all in.

Option 3: Use ERPClaw, free. This is the option we built. ERPClaw is AI-native (the assistant is the primary interface) and runs Shopify COGS tracking automatically for every order, per warehouse, with FIFO valuation, and posts the journal entries straight into a built-in general ledger. No middleware, no QuickBooks subscription, no premium tier paywall. The full feature list is on our Shopify integration page, and pricing is documented (it is free, forever, open source) on our pricing page.

Here is what ERPClaw actually does for Shopify cost of goods sold out of the box:

- Watches every Shopify order webhook in real time and posts COGS at the moment of sale, not at month-end

- Tracks cost per warehouse, so a Toronto-shipped mug carries its $4 cost and an Atlanta-shipped mug carries its $5.20 cost into the same gross-margin report

- Uses FIFO by default, with a one-line config switch to weighted average if your accountant prefers it

- Reverses COGS automatically when a refund comes through, including partial refunds

- Lets you bake inbound freight and duty into landed cost so the field on the order shows the real number, not just the wholesale invoice price

- Publishes every journal-entry rule openly (we list the GL posting rules on the Shopify features page so you can see the math, not just the answer)

If you want to see how the install flow works on a real store, the Shopify install walkthrough is a five-minute walk-through. If you are evaluating against the broader ecommerce inventory cost stack, the for-ecommerce overview covers the rest of the picture: gift cards, payouts, returns, the lot.

## Shopify Cost of Goods Sold FAQ

Does Shopify automatically calculate cost of goods sold?

Partially. Shopify uses the “Cost per item” field you enter on each product variant to compute a basic profit number in Shopify Analytics. It does not include inbound freight, duties, payment processing fees, or returns, and it does not handle multi-warehouse cost differences. For an accurate Shopify cost of goods sold number you need either a manual spreadsheet, a paid accounting app, or an ERP like ERPClaw that tracks landed cost and per-warehouse FIFO automatically.

What is the difference between cost of goods sold and operating expenses?

COGS is the cost directly tied to the units you sold (wholesale, freight in, duty, packaging). Operating expenses are the costs of running the business that exist whether you sell zero units or a million (rent, software, marketing, salaries). Both reduce your profit, but only COGS reduces your gross margin. A bank or buyer will look at gross margin to judge how strong the underlying product economics are, which is why getting COGS right matters more than getting operating expenses right.

Should shipping be included in cost of goods sold?

Inbound freight (getting product into your warehouse) is always COGS. Outbound shipping (sending it to the customer) depends on how you structure pricing. If you charge for shipping, most accountants put both the revenue and the cost in COGS. If you offer free shipping baked into the price, the cost has to come out of margin somewhere, and putting it in COGS gives you the most honest unit economics.

How do I calculate COGS Shopify when I have two warehouses with different costs?

You need a system that tracks cost at the inventory level, not just the variant level. Shopify shows one Cost per item per variant, so you cannot do this in Shopify alone. ERPClaw stores cost per warehouse and applies FIFO at the warehouse level, so an order shipped from your East Coast 3PL carries that warehouse’s cost and a West Coast order carries its own.

What is FIFO and do I have to use it for my Shopify store?

FIFO means “first in, first out.” Say you bought one batch of mugs at $4 and a later batch at $5.20, FIFO says the next mug you sell came from the $4 batch. The IRS allows FIFO, weighted average, and a few other methods. You pick one, you stick with it, and you tell your accountant. Most Shopify stores find FIFO matches reality the closest because inventory genuinely flows oldest-first off the shelf.

Does ERPClaw really do COGS tracking for free?

Yes. ERPClaw is open-source under open source license. Per-warehouse FIFO COGS tracking is in the free tier; there is no premium tier. Compare that to A2X, which gates COGS to its $49+/month plan, on our A2X comparison page. The catch is that ERPClaw is self-hosted, so you run it on your own machine or server. If you want managed hosting with backups handled for you, ERPClaw Cloud Managed is our hosted offering, onboarding by request through /pricing/; the open-source self-hosted version stays free forever.

How does ERPClaw handle returns and refunded COGS?

When Shopify sends the refund webhook, ERPClaw automatically reverses the original journal entry: credit COGS, debit inventory, exactly the inverse of the sale. Partial refunds reverse pro-rata. The result: your gross margin number always reflects net sales after returns, not gross sales, which is the number that actually matches what hit your bank account.

## The Honest Closing Pitch

Shopify gives you a starting point. It does not give you the real Shopify cost of goods sold number, and the gap between the product-page field and your bank statement is where most stores lose visibility into actual unit economics. You can close it with a spreadsheet (slow), a paid app stack (expensive), or an open-source ERP that does the math for you (free).

ERPClaw installs in five minutes and connects to Shopify with one OAuth click. It tracks COGS per warehouse with FIFO valuation, posts the GL entries, and reverses them when refunds come back. No premium tier, no usage cap, no QuickBooks bill. Open source on GitHub. The pricing page is short because the price is zero.

Start with the Shopify install walkthrough, or read the full breakdown on our Shopify integration page. Whatever tool you pick, stop running ads against the wrong margin number. The bank account already knows.

---

## Double-Entry Bookkeeping for Developers, in Code
URL: https://www.erpclaw.ai/blog/double-entry-bookkeeping-for-developers/

# Double-Entry Bookkeeping for Developers, in Code
Double-entry bookkeeping explained the way developers think: invariants, immutable logs, and why debits equal credits, with code.

Most developers understand database transactions, append-only logs, and immutable event streams. Most developers also bounce off accounting tutorials within ten minutes because the tutorials open with “debits go on the left and credits go on the right” and never recover.

This is the post I wish someone had handed me before I started building ERPClaw. It is double entry bookkeeping explained as a system a developer would design, not as a 1494 manuscript by an Italian Franciscan monk. By the end you should understand the entire model well enough to read a journal entry, post one, and recognize when an ERP is doing it wrong.

This post is not for accountants. Accountants will find it shallow. It is for developers who need a working mental model in 20 minutes so they can ship code that does not corrupt financial data.

## The mental model in one sentence

Double entry bookkeeping is an append-only log where every entry is a balanced pair of writes against two or more accounts, where the sum of writes always equals zero.

That is it. Everything else is naming conventions, sign conventions, and the rules people built on top.

## The five account types as data buckets

An account is a labeled bucket that holds a running balance. There are five canonical types. Forget what they are called for a second and look at what they actually represent in a system you would design.

| Account type
| What it represents
| Example

| Assets
| Things the company owns
| Cash, Inventory, Accounts Receivable

| Liabilities
| Things the company owes
| Loans, Accounts Payable, Deferred Revenue

| Equity
| What the owners have invested or earned
| Retained Earnings, Owner Capital

| Revenue
| Money the company earned (top line)
| Product Sales, Subscription Revenue

| Expenses
| Money the company spent
| Rent, Salaries, Cost of Goods Sold

Every account in a chart of accounts is one of these five types. ERPClaw ships with a sensible default chart of accounts you can extend; other tools ship more or fewer, or let you create as many as you want. Doesn’t matter, every one of them is one of these five types underneath.

The fundamental equation that ties them together:

`Assets = Liabilities + Equity`
Or if you split equity into its components:

`Assets = Liabilities + (Owner Capital + Retained Earnings)
Retained Earnings = Cumulative Revenue - Cumulative Expenses`
So really:

`Assets = Liabilities + Owner Capital + Cumulative Revenue - Cumulative Expenses`
This equation must always hold. Every transaction must preserve it. That is the invariant a double entry system enforces.

## Why two writes per transaction

Imagine you record only one write per transaction. A customer pays you $100. You write `Cash += 100`. The cash bucket goes up. Done.

Now what? Where did the $100 come from? You have no idea. You broke the accounting equation. Assets went up by $100, but liabilities and equity did not change. The books no longer balance. You cannot generate a balance sheet because the balance sheet would not balance.

The fix is to require every transaction to record both sides of what happened. The customer paid you $100, which means cash went up AND something else changed. Maybe revenue went up (you sold them a thing). Maybe a liability went down (they paid off a bill). Maybe deferred revenue went up (they prepaid for a subscription). Whatever it is, you record it explicitly.

`Transaction: customer pays $100 for a product
Cash += 100 (asset goes up)
Revenue += 100 (revenue goes up, which flows into equity via retained earnings)`
Now the books balance. Assets went up by $100. Equity (via cumulative revenue) went up by $100. The equation holds.

This is double entry. Every transaction is two or more writes that together preserve the equation.

## Debits and credits, finally

Now we can talk about debits and credits. The accounting world uses these two words because writing `+=` and `-=` against five different account types gets confusing fast (some go up when you “add” and some go down when you “add” depending on context). Debit and credit are sign conventions that paper over the confusion.

Here are the only rules you need:

Debit = left column. Credit = right column. That is the visual convention.

For the five account types, debit and credit have these effects:

| Account type
| Debit
| Credit

| Asset
| increases
| decreases

| Liability
| decreases
| increases

| Equity
| decreases
| increases

| Revenue
| decreases
| increases

| Expense
| increases
| decreases

In every transaction, the total debits must equal the total credits. That is the rule.

Why are the signs flipped between asset and liability? Because of the equation. Assets on the left of the equation, liabilities on the right. Pushing the same number to both sides keeps the equation balanced. Debit increases the left side, credit increases the right side. The convention is just sign management.

The example transaction in debit/credit form:

`Transaction: customer pays $100 for a product
DR Cash 100 (asset goes up by 100, so debit)
CR Revenue 100 (revenue goes up by 100, so credit)`
Total debits: 100. Total credits: 100. Balanced.

Once you internalize the table above, every journal entry you ever read makes sense in about three seconds. The brain trick is that “debit” is just the technical word for “this side of the entry” and the effect on the balance depends on the account type.

## A journal entry is a transaction in the database sense

In code, a journal entry is roughly this:

`@dataclass
class JournalEntryLine:
account: str
debit: Decimal
credit: Decimal
# exactly one of debit/credit is non-zero per line

@dataclass
class JournalEntry:
id: str # UUID
posting_date: date
description: str
lines: list[JournalEntryLine]

def is_balanced(self) -> bool:
total_debit = sum(line.debit for line in self.lines)
total_credit = sum(line.credit for line in self.lines)
return total_debit == total_credit

def post(self, db):
if not self.is_balanced():
raise ValueError("Journal entry must balance")
with db.transaction():
for line in self.lines:
db.execute(
"INSERT INTO gl_entry (account, debit, credit, je_id) VALUES (?, ?, ?, ?)",
line.account, line.debit, line.credit, self.id,
)`
That is the actual data model. A journal entry is a header plus N lines. Each line debits or credits one account. The entry is valid if total debits equal total credits. Posting it is one DB transaction that inserts all the lines atomically.

In ERPClaw, this is exactly how the `gl_entry` table works. Every line is one row. Every journal entry is N rows tied together by `je_id`. The constraint that debits equal credits is enforced at posting time. If a posting tries to write an unbalanced entry, the transaction rolls back and nothing hits the table.

## Append-only and immutable, like a transaction log

Here is the part developers find genuinely interesting. The general ledger (GL) is append-only and immutable. You never update a posted entry. You never delete one. The same way a Kafka log or a git commit history works, the GL is a write-once record of what happened.

If you posted an entry and you need to “fix” it, you do not edit the original. You post a new entry that reverses the original (mirrors the debits and credits) and then post a corrected entry. The audit trail shows both: the original posting, the reversal, the correction. Nothing is hidden.

This matters because the GL is what your auditor reads. If posted entries could be edited or deleted, there would be no way to prove the books were not tampered with. The append-only constraint is the root of trust in financial accounting.

In ERPClaw, the `gl_entry` table has no `updated_at` column on purpose. There is no UPDATE statement anywhere in the codebase that targets it. There is no DELETE either. The only way to “remove” an entry is to post a reversal, which is a new entry with flipped debits and credits.

This is also why most accounting bugs in custom software are so painful. Developers come from environments where mutability is normal. They write `UPDATE invoices SET amount = ... WHERE id = ...` and the books fall apart silently. The fix is architectural: the GL is a log, not a state.

## The income statement and balance sheet are derived state

Once you understand the GL is the source of truth and it is append-only, the financial reports are easy.

Trial balance. Group every entry by account, sum debits, sum credits. The balance per account is debits minus credits, signed by account type.

`def trial_balance(db, as_of: date) -> dict[str, Decimal]:
rows = db.query(
"SELECT account, SUM(debit) AS d, SUM(credit) AS c "
"FROM gl_entry WHERE posting_date <= ? GROUP BY account",
as_of,
)
return {r.account: r.d - r.c for r in rows}`
The trial balance must sum to zero across all accounts. If it does not, you have a bug.

Balance sheet. Take the trial balance, partition by account type, sum assets, liabilities, equity. Display.

Income statement. Take the trial balance, partition into revenue and expense accounts only, take credits minus debits for revenue and debits minus credits for expense, compute net income.

Cash flow statement. A bit more involved (you need to derive operating, investing, financing activities), but again derivable from the GL plus a category mapping on each account.

The financial reports are not stored. They are queries against the GL. Every report is reproducible from the same log. This is why accountants can sleep at night.

## What the 12 step GL validation actually checks

ERPClaw runs every GL posting through a 12 step validation pipeline before it commits. Every step is a check that prevents a class of bug. For developers, here are the checks worth knowing about:

- All amounts are positive Decimals. No negative amounts anywhere. Reversals are handled by flipping debit and credit, not by negating the amount.

- Each line has exactly one of debit or credit, not both. A line that debits and credits the same account is meaningless.

- Total debits equal total credits. The fundamental balance check. Floats fail this. Decimals do not.

- At least two distinct accounts. A “transaction” that debits and credits the same account is a no-op.

- Posting date within an open period. You cannot post into a closed accounting period.

- Account exists and is not archived. No phantom accounts.

- Currency consistent within the entry. All lines in the same currency. Multi-currency is a separate concept handled with FX entries.

- Reference document exists if claimed. If the entry references invoice ABC, invoice ABC must exist.

- Reference document is not already posted. No double posting.

- Idempotency key is unique. The same operation cannot post twice if retried.

- All numeric values are Decimal, not float. Floats lose pennies. Decimals do not.

- GL state remains balanced after posting. Trial balance still sums to zero.

Any one failure rolls back the whole transaction. Nothing partial gets written.

This is what makes the GL trustworthy. The constraints are explicit, enforced at write time, and tested by the L0 constitutional test suite which catches regressions before they ship.

## Why floats are a bug

In most ERPClaw code, money is stored as TEXT in SQLite and represented as Python `Decimal` in memory. Never as `float`. Never as `int` of cents.

The reason is the IEEE 754 floating point representation cannot exactly represent most decimal values. `0.1 + 0.2 == 0.3` is False in Python because both `0.1` and `0.2` are approximations. The error is small per operation but accumulates. Across thousands of GL entries, you end up with a trial balance that “almost” sums to zero, with a sub-penny error that is impossible to track down.

`Decimal` represents the exact decimal value. `Decimal('0.1') + Decimal('0.2') == Decimal('0.3')` is True. Storing as TEXT in SQLite preserves the string representation losslessly. This is non-negotiable.

`from decimal import Decimal, ROUND_HALF_UP

amount = Decimal('100.05')
tax = (amount * Decimal('0.0825')).quantize(Decimal('0.01'), rounding=ROUND_HALF_UP)
# tax = Decimal('8.25')
# total = amount + tax = Decimal('108.30'), exactly`
Any financial code that uses `float` is wrong. Any database schema that stores money as REAL is wrong. The constraint is mechanical and tested by ERPClaw’s L0 suite (test 023: “no float in money paths”).

## What submit means in a draft to submit lifecycle

Most ERP entities have two lifecycle states: draft and submitted. The draft is editable. The submitted version is locked, posts to the GL, and triggers downstream actions.

The pattern in code:

`def add_invoice(...) -> str:
"""Create a draft invoice. No GL posting. Editable."""
invoice_id = str(uuid.uuid4())
db.execute("INSERT INTO sales_invoice (id, status, ...) VALUES (?, 'draft', ...)", invoice_id, ...)
return invoice_id

def submit_invoice(invoice_id: str):
"""Validate, post GL, lock the invoice. Single transaction."""
with db.transaction():
invoice = load_invoice(invoice_id)
validate_for_submit(invoice)
gl_entry = build_gl_entry(invoice)
run_12_step_validation(gl_entry)
post_gl(gl_entry)
update_inventory(invoice)
update_ar_aging(invoice)
db.execute("UPDATE sales_invoice SET status = 'submitted' WHERE id = ?", invoice_id)`
The whole submit operation is one DB transaction. Any failure rolls back everything. The invoice does not become submitted unless every cross-table write succeeds. There is no partial state.

This is the data integrity guarantee. Either the invoice is fully submitted with GL, inventory, and AR updated, or none of it happened. No “the invoice is submitted but the GL did not post.” That state is unreachable.

## The whole thing in 100 words

Double entry bookkeeping is an append-only log of balanced transactions against five account types (assets, liabilities, equity, revenue, expenses) where every transaction preserves `Assets = Liabilities + Equity + Revenue - Expenses`. Each transaction is two or more writes that sum to zero in debit/credit terms. The GL is immutable; corrections are reversals plus new entries. Reports are derived state, not stored. Money is `Decimal`, not `float`. Every posting is one DB transaction that either fully commits or fully rolls back. That is the model. Everything else is convention or jargon.

## Where to read more

If you want the actual code, ERPClaw is open source. The GL posting code lives in `erpclaw_lib/gl/` in the shared library. The 12 step validation lives in `validate_gl.py`. The L0 constitutional tests live in `testing/l0/`.

For the conceptual reading, the AICPA’s free guides are decent and short. Ignore textbooks. They are written for accounting students, not developers, and they bury this 100 word model in 600 pages of jargon.

## CTA

If you are building software that touches financial data and you want to see how a working open-source ERP implements the constraints in this post, ERPClaw is open source, AI-native, and runs on SQLite or PostgreSQL via PyPika.

- Install ERPClaw (5 minutes)

- Browse the source on GitHub

- SQLite or PostgreSQL: the database choice

- See the engineering blog

- Run the demo

The full back-office suite is free, the GL implementation is open, and the test suite that enforces all of the above is on GitHub. If you find a bug in the GL validation, file an issue. If you have a 13th check we should add, send a pull request.

## FAQ

### Do I really need to use Decimal everywhere or can I use int cents?

Int cents (storing $100.05 as `10005`) works for currencies with two decimal places and no rounding edge cases. It breaks the moment you compute tax (which produces fractional cents that need to round) or apply a percentage discount. Decimal handles all of these correctly with explicit rounding rules. Use Decimal.

### Why is the GL append-only instead of using soft deletes or version columns?

Soft deletes and version columns require a query that filters on “the latest version” or “not deleted” to read the current state. That filter is easy to forget. The append-only constraint with no UPDATE or DELETE is enforced by the database schema itself, not by query convention. It is impossible to accidentally read stale data.

### How do I undo a posted journal entry without violating immutability?

Post a reversal. The reversal is a new journal entry that mirrors the original (debits become credits and vice versa). The original posting plus the reversal nets to zero. If you also need to post a corrected entry, post that as a third entry. The audit trail shows all three.

### What about multi-currency? Do amounts still need to balance?

Yes. Multi-currency entries balance in their original currency and also in the reporting currency, which means each line carries an FX rate at the time of posting. The reporting currency totals are derived. Currency mismatches across lines (e.g., one line in EUR and another in USD) require an explicit FX clearing account to balance. ERPClaw ships 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED) on the payment side: an invoice settles in the currency it was raised in, and no FX conversion ever happens inside a transaction.

### Is double entry overkill for a one-person business?

It depends on whether you ever want to take the business seriously. If you ever need a bank loan, raise capital, sell the business, or get audited, you need double entry. Most personal finance tools are single entry (cash in, cash out) and that is fine for personal use. The moment a business has receivables, payables, inventory, or deferred revenue, single entry stops being able to represent reality.

### Can I implement this in Postgres or do I need SQLite?

Either works. ERPClaw is database-agnostic via PyPika and supports both. The append-only constraint is enforced in application code; the schema only needs to support transactions and unique constraints. Most relational databases qualify.

### Why is “credit” sometimes positive and sometimes negative depending on the report?

Different reports flip the sign convention to make numbers look intuitive. On the income statement, credit revenue is shown as positive (because revenue is “good”). On the balance sheet, credit liabilities are shown as positive (because they are an obligation). Internally the GL stores debit and credit as separate columns. The report layer applies the sign convention. The internal data is unambiguous.

Install ERPClaw and read the GL code if you want to see this implemented in production-grade open source.

---

## ERPClaw 4.15: Tested the Way You Actually Work
URL: https://www.erpclaw.ai/blog/erpclaw-4-15-tested-the-way-you-work/

# ERPClaw 4.15: Tested the Way You Actually Work
ERPClaw 4.15 is live: every release now clears complete business scenarios checked against the real books, and short payments are written off in the payment.

Version 4.15 is live on every channel. There is no new screen to show you, and
that is rather the point. This release is about money landing where it should,
and about being able to prove that it did. Three parts of it are worth your
time.

## We test it the way you would actually use it

Software is usually tested by exercising every button. That tells you the
buttons work. It does not tell you your books are right.

Before this release could ship, it had to get through a set of complete business
scenarios, each one driven in plain business language from beginning to end: a
customer short-pays an invoice, a recurring billing run turns into a real
invoice, an invoice raised outside the billing flow gets attached to the period
it belongs to, an order runs all the way to cash with corrections along the way,
a deposit is forfeited and kept. Then a deterministic check reads the books that
came out of the other end and decides pass or fail on the numbers, not on the
screens.

Every scenario passed on the code we shipped, and the books-integrity rules held
in all of them. That gate is the floor now. A scenario that ends with the wrong
balance stops the release, whatever the buttons did.

The reason to care is that those scenarios are your Tuesday. They are not
synthetic edge cases invented to keep a test suite busy. They are the awkward,
ordinary things that happen in a real business and quietly break accounting
software.

## Write-offs, the way an accountant would do them

Two situations, both of which used to be tidier in theory than in practice.

The first is the short payment. Say a customer owes you $4,000, pays $3,850, and
considers the matter closed. What you want is to record what they actually paid
and write off the difference in the same movement, so nothing is left dangling
on an invoice that will never be settled. That is now one step, taken as part of
the payment itself:

“Record their payment against this invoice and write off the short amount.”

The second is bad debt. When you decide an invoice is not going to be paid,
ERPClaw posts balanced bad-debt entries against that invoice, dated by the
decision rather than by the original sale. The distinction is not pedantry. The
loss belongs to the period in which you made the call, and quietly backdating it
into a period you have already closed is how a clean set of books turns into an
awkward conversation. Both write-off routes ask you to confirm before anything
moves, and in the legal edition, writing off an invoice now moves the actual
books instead of only changing a status.

## Corrections you can see

If you run more than one company, the consolidation step that cancels out
trading between them is now safe to run twice. It eliminates only what is new,
so a second run does not double the adjustment. If an earlier repeat run left
damage behind, the consolidated trial balance names the leftover rows instead of
absorbing them quietly, a listing shows you exactly which rows they are, and
removing them takes an explicit confirmation and records every removal.

Upgrades got the same treatment. A migration now declares whether it touches
your data, and the ones that do write an audit-trail row for every document they
change, inside the same transaction as the change itself. When an upgrade
repairs something in your history, you can see precisely what it repaired.

Two shortcuts were retired rather than patched: a legacy elimination pair that
posted group adjustments into operating companies’ own books and left them
unbalanced, and two stock actions that could move quantity without a matching
entry in the books. Both now decline and name the flow that does the job
properly. A closed route is better than a quietly wrong one.

Underneath all of it sits a new always-on rule: every customer’s payment history
has to agree with their open invoices minus any unapplied cash, exact to the
cent, at all times. Around that, a run of repairs. Cancelling an invoice now
releases the cash that was applied to it, so money stops being held against a
document that has left the books. What a customer owes is no longer counted
twice. Loan money reaches the books instead of vanishing into a swallowed error.
A disposed asset posts its proceeds and its gain or loss to the right accounts
rather than to depreciation. This is the same argument as our
audit-ready books page, taken one step further.

## PostgreSQL, verified

The supporting technical note. Every schema change, connection and catalog
lookup now runs through a single seam in the code, and the module install path
is verified against a live PostgreSQL server rather than against SQLite alone.
PostgreSQL or SQLite remains your choice, the same software either way, and the
reasoning behind that choice is written up in
SQLite or PostgreSQL. MySQL is
not supported, because the schema cannot be indexed properly there, and it now
says so plainly instead of half-working.

## FAQ

What happens when a customer pays less than the invoice?

Record the payment and write off the difference in the same step. The shortfall
posts as a real bad-debt entry instead of sitting on the invoice as a balance
nobody is ever going to collect.

Should a bad debt be dated to the sale or to the decision?

To the decision. ERPClaw dates a write-off by the day you decide the money is
not coming, so the loss lands in the period where you made the call rather than
in one you have already closed.

Do I need to do anything special to upgrade?

Run the upgrade. Where existing data needs repairing, the migrations do the
repair and record what they changed, document by document, so nothing about the
fix is invisible to you.

Can I run the consolidation more than once?

Yes. Elimination generation only eliminates new activity, so a repeat run does
not double anything, and whatever an older repeat run left behind is reported
to you and can be removed with an explicit confirmation.

What does this cost?

Nothing. ERPClaw is free and open source, self-hosted, and every
part of this release is in it.

Tagsreleaseaccountingwrite-offsauditopen-source
Share:XLinkedInHacker News

---

## ERPClaw's Shopify Integration: Free and Live Today
URL: https://www.erpclaw.ai/blog/erpclaw-now-on-shopify-app-store/

# ERPClaw's Shopify Integration: Free and Live Today
ERPClaw's Shopify integration is free, open source, and installable today: orders, refunds and payouts posted into real books.

Today I am shipping the Shopify integration for ERPClaw. The public
App Store listing is still pending review, but the integration itself
is installable today. It is free, open source, and replaces what apps
like A2X, Synder, and Bookkeep charge a recurring monthly fee to do.

The integration syncs every order, refund, payout, and dispute from
your Shopify store directly into your ERPClaw general ledger. It posts
a journal entry per transaction, with a three-layer payout
reconciliation that either hits zero or tells you exactly why it did
not.

If you run a Shopify store and you have ever spent a Saturday matching
Shopify payouts to your bank deposits in a spreadsheet, this post is
for you.

## What you get

Two-click pairing from the Shopify Admin. The app issues a six-character
pairing code, you run `erpclaw shopify-connect --pairing-code ABC-XYZ`
on your own ERPClaw instance, and you are done. Behind the scenes:

- The full set of GL accounts auto-created (clearing, revenue, shipping revenue, tax payable, COGS, inventory, fees, discount, refund, chargeback, chargeback fee, gift card liability, reserve, bank)

- Full historical sync of orders, refunds, payouts, and disputes

- Real-time webhook subscription so new events post within seconds

- Status push back to the Shopify Admin so it shows your sync health

- Cross-platform daemon (launchd on macOS, systemd-user on Linux, cron fallback) so you never have to think about it again

## The clearing account pattern

Every order you receive goes through a clearing account on the way to
your bank. ERPClaw posts the gross sale to revenue, the Shopify fees
to expense, and the net to clearing. When Shopify pays you out three
days later, ERPClaw posts the bank deposit and reduces the clearing
account by the same amount. If the clearing account ever fails to
zero, something is broken and you will know immediately.

Most paid Shopify accounting apps post summary journals at the end of
the day. That hides the per-transaction detail you need for refunds,
disputes, and audits. ERPClaw posts a journal entry per transaction
with full Shopify metadata attached. Slower? Yes, by a few seconds.
Worth it? Every audit you ever survive.

## What the paid apps charge for

The paid apps gate the features serious merchants actually need behind
their higher tiers: COGS tracking, multi-location inventory, gift card
deferred revenue, dispute reserves. (Current pricing is on each
vendor’s site: A2X,
Synder, Bookkeep.)

ERPClaw includes all of that for free. Not as a freemium tier with
the good stuff locked behind a paywall. Free as in open source, free
as in source code on GitHub, free as in self-hosted on your own
infrastructure.

## What it does not do

This is not a replacement for Shopify itself. We do not handle the
storefront, payment processing, or merchant of record. Shopify does
all of that, and ERPClaw plugs into the back of it.

This also is not an enterprise ERP killer. If you are running SAP at
ten thousand stores, ERPClaw is not for you. We are aiming at the
small and mid-market merchants who have been priced out of proper
accounting tooling and are running their books in a spreadsheet
because QuickBooks does not understand Shopify payouts.

## How to get started

The public App Store listing is still pending review, but you do not
have to wait for it. Start at the
Shopify integration page for the feature tour,
then follow the setup docs to connect your store. The
pairing code shows up after the OAuth handshake. Run the connect
command on your own ERPClaw instance and you are live.

If you have not installed ERPClaw yet, the full setup is documented
at erpclaw.ai/docs. Five minutes from
zero to a working ERP, no consultants required.

For merchants who want to skip the App Store entirely (air-gapped
installs, custom-app power users, paranoid security teams), there is
a custom-app flow: create a Shopify Custom App in your Partners
dashboard, grab the `shpat_` access token, and run
`shopify-add-account` with that token. Skips the Worker, skips the
status push, gives you complete control.

## Why we built this

Our co-founders spent years implementing enterprise ERP rollouts and shipping operations software. We know exactly which parts of a million-dollar ERP are genuinely complex and which parts are just expensive. Shopify accounting is in
the second category.

The math behind double-entry bookkeeping has not changed since 1494.
The work in a Shopify accounting app is plumbing: read the API,
parse the schema, post to a GL, reconcile the payout. None of it
justifies a recurring monthly fee, and certainly not for the
merchants who need it most.

ERPClaw is what happens when you write the spec precisely, hand it
to Claude, and refuse to charge for commodity business logic.

## What is next

The Stripe integration is already live (ASC 606 revenue recognition,
Connect platform fees, MRR reports). Both run on the same ERPClaw
core, share the same chart of accounts, and reconcile against the
same bank deposits.

We are watching for App Store review feedback and any merchant
issues that turn up as more stores connect. Star the GitHub repo if
you want to follow along, and email `support@avansaber.com` if
anything breaks.

The integration is live now. Welcome to the new economics of
Shopify accounting.

---

## ERPClaw Q2 2026: What Shipped, What's Next
URL: https://www.erpclaw.ai/blog/erpclaw-q2-2026-state-of-the-platform/

# ERPClaw Q2 2026: What Shipped, What's Next
The first quarterly state of the platform. Two deep integrations live, 73 pages of marketing site, a rebuilt persona-driven demo, and what is next.

This is the first ERPClaw quarterly update. The headline for Q2 2026: we shipped two deep integrations, deployed seventy-three pages of marketing site, and rebuilt the demo around three personas with fourteen scenarios.

I am going to tell you what shipped, what is broken or partial, what is coming next, and the actual numbers behind the project. No spin. The point of writing these every quarter is to leave a public paper trail that an investor, an advisor, a potential user, or a future me can scroll through and tell whether the curve is bending the right way.

If you have not seen ERPClaw before, the short version: it is an open source, AI-native ERP. It covers the full back office (sales, purchasing, stock, manufacturing, HR, payroll, projects, and more) with a deep, scriptable action surface across the whole system. One SQLite database that lives on disk in a single file you own. open source, free, runs locally, driven by your AI assistant in natural language. The full story of why our co-founder Nikhil started building it lives in from SAP consultant to open source builder.

Now to this quarter.

## What shipped

### Stripe v1.0 went live on Stripe Marketplace

The biggest shipment of Q2 was the Stripe deep integration reaching v1.0 and getting listed on the Stripe Marketplace. Deep Stripe coverage: payments, payouts, fees, disputes, subscriptions, full ASC 606 revenue recognition, MRR and churn reports, Stripe Connect application fee accounting, and per-event GL posting that survives the twelve step validation engine.

The internal architecture is the part I am most proud of. Every Stripe webhook lands in an event store, gets deduplicated by signature plus event ID, fans out to a handler that writes draft entries, and then either auto-submits or queues for review based on the merchant’s configured policy. Connect fees, refunds, disputes, payouts, and subscription proration all flow through the same pipeline. The accountant sees journal entries with the original Stripe object IDs in the memo line so any line item is traceable back to the source in two clicks.

The full feature surface lives at /features/stripe and the integration docs are at /docs/stripe/install-walkthrough/. If you want the deep dive on why ASC 606 belongs inside the integration rather than in a separate report tool, that is at /blog/stripe-asc-606-revenue-recognition-guide/.

For context on where this lands in the market, the Rillet alternative comparison is the cleanest framing. Rillet charges around $1,500 a month for a similar surface. ERPClaw charges $0.

### Shopify v1.1.3 with App Store OAuth, status mirror, GDPR webhooks, and a daemon

Shopify caught up to Stripe in code surface this quarter. v1.1.3 ships with deep order, payout, and reconciliation coverage matching the Stripe side, and a real installation flow that goes through a Cloudflare Worker for OAuth pairing rather than the v1 hand-pasted token mess that almost cost me a week.

The new bits in v1.1.3:

- App Store OAuth flow via Cloudflare Worker (the Worker holds nothing; it does the token exchange and hands the credential back to the local install)

- Status mirror so the merchant can see installation health, last sync, webhook backlog, and reconciliation status from a single page

- All seven mandatory GDPR webhooks (customers/data_request, customers/redact, shop/redact, plus the four newer privacy events)

- Daemon mode that subscribes to the Shopify event bus and processes webhooks asynchronously rather than the v1 polling loop

App Store submission is open but pending review. Shopify’s queue is what it is. I will write a separate post when the listing goes live.

The Shopify launch post is at /blog/erpclaw-now-on-shopify-app-store/ and the full feature page is at /features/shopify. For the buyer-intent comparison, the A2X alternative post and the Synder alternative post cover the head-to-head.

### Website rebuild from twenty-eight to seventy-three pages

The marketing surface at start of quarter was twenty-eight pages, mostly placeholder copy left over from the launch. By end of quarter it is seventy-three pages, all real, all indexed, all built on Astro content collections so we can scale without architectural drag.

The breakdown:

- 9 comparison pages (A2X, Synder, Rillet, Bookkeep, Webgility, Xero, FreshBooks, ERPNext, QuickBooks)

- 13 vertical pages (ecommerce, SaaS, agency, restaurant, manufacturer, nonprofit, legal, real estate, solo founder, small business, accountants, healthcare, education)

- 12 documentation pages (six per integration, plus the landing matrix)

- 17 blog posts (this one is the eighteenth)

- The remaining headline pages: features, pricing, demo, security, quality, privacy, the integration landings, and the founder page

Schema.org coverage went from one (incorrectly applied SoftwareApplication on every page) to eight correctly typed schemas (SoftwareApplication for the homepage, Product for feature and comparison pages, FAQPage on pricing, BlogPosting on every blog, Article on every doc, BreadcrumbList auto-derived on every non-home page, Organization on the footer, AboutPage on the founder and security pages). Sitemap went from a 404 to fifty-eight indexed URLs (and counting; we add as we publish).

If you want to inspect the catalog yourself, /docs is the integration matrix and /blog is the full post index.

### Animated demo with three personas and fourteen scenarios

The old demo was a single linear walkthrough that showed a generic small business. It converted poorly because nobody is a generic small business. They are a Shopify merchant, or a SaaS founder, or an agency owner, or a restaurant operator, and they want to see the screens that match their reality.

The new /demo ships with three personas (Shopify merchant, SaaS founder using Stripe, multi-location restaurant operator) and fourteen branching scenarios across them. Each scenario is a real flow with real data running against a real ERPClaw instance, animated so a visitor can see what the assistant does without having to install anything. The state persists in localStorage so a visitor can leave and come back to where they were.

This is Phase 1 of the demo plan. Phase 2 is the in-browser sandbox using WebVM/CheerpX so a visitor can actually drive the system rather than watch it. That is queued for Q3 once the commercial license review closes.

### Analytics, AI bot opt-in, and per-page OG image generation

A handful of smaller things that mattered:

- GA4 installed in the Layout component so every page emits engagement, time on page, and conversion events

- robots.txt rewritten to explicitly allow ClaudeBot, GPTBot, PerplexityBot, Google-Extended, Bytespider, CCBot, Applebot-Extended, Amazonbot, and meta-externalagent (the bet is that AI search citations are going to matter more than rank position over the next eighteen months, and the right move is to be in the training data on day one)

- Per-page OG image generation so every comparison, vertical, doc, and blog post gets a branded Open Graph card automatically when the page builds

### Founder writeup picked up by HackerNoon

The HackerNoon feature on the previous indie SaaS project (ZapInventory) ran in mid-Q2. It drove a small but real spike in GitHub stars and inbound LinkedIn messages. The follow-up post on ERPClaw specifically is being drafted for Q3 placement.

## What is broken or partial

I would rather flag this stuff myself than wait for a user to find it.

### Shopify App Store listing is pending

The integration is built, tested, and submitted. The listing is in Shopify’s review queue. I do not control the timeline. Until it lands, merchants can install via the custom app flow documented at /docs/shopify/app-store-vs-custom-app/, which is functionally identical but does not get the App Store distribution.

### Multi-currency support gap

Both Stripe and Shopify integrations currently assume a single base currency. If your store sells in EUR and your books are in USD, you have to manually configure the exchange rate per period. This is the single largest gap I am hearing about from prospective users. Multi-currency v1.x is the highest-priority Q3 item, scoped to support FX revaluation, presentation currency separate from functional currency, and intercompany consolidation for multi-entity setups.

### WebClaw UI for Stripe and Shopify is still CLI-only

WebClaw, the optional dashboard layer, currently does not surface the Stripe or Shopify integration views. Everything works via the natural language assistant or direct CLI invocation, but a merchant who wants a clickable dashboard for their Stripe MRR or Shopify reconciliation status has to wait. The dashboard views are in flight; ETA is mid Q3.

### Main website server deploy backlog

The avansaber.com main website (the parent company site, not erpclaw.ai) has a privacy-stripe.php page authored locally but blocked on an SSH key issue with the GitLab origin. The Shopify privacy page is live; the Stripe one is stuck in commit limbo. Cosmetic for now since the integration itself does not depend on it, but I want it cleared before pushing the App Store listing live.

### Server deploy of the seventy-three page site happened post-quarter-close

The full Astro build finished on 2026-04-26 and the rsync to RunCloud landed shortly after. Traffic numbers below reflect that, which is to say: the site has been live in its current shape for less than a week as of this writing. Q3’s quarterly post will have the first real traffic curve.

## What is next

The Q3 priorities in order:

### App Store approval and the public launch beat

When Shopify’s App Store listing goes live, that triggers a coordinated launch sequence: HackerNoon op-ed, ProductHunt, HN, the Shopify subreddit, LinkedIn, X. The launch blog draft is already written and sitting in apps/shopify/LAUNCH_BLOG_DRAFT.md. The countdown starts the day Shopify approves.

Stripe Marketplace is already live so the Stripe push is more of a sustained content drumbeat than a single launch event.

### WebClaw dashboard views for Stripe and Shopify

Pulling the integration data into clickable dashboard views. The CLI surface stays as the source of truth; the dashboard is a thin read layer on top. Target ship: mid Q3.

### Multi-currency v1.x

The biggest functional gap. Q3 scope: FX rate management, period revaluation, presentation versus functional currency, and intercompany consolidation for the multi-entity case. This is a multi-week build and will get its own deep dive post when it ships.

### Vertical content build-out

Of the ninety-five posts mapped in the blog universe, seventeen are live. The Q3 plan is twenty-six more posts focused on comparison and education content, plus the first batch of vertical guides (Shopify store accounting, SaaS accounting with Stripe ASC 606, agency time billing, restaurant prime cost). The point is to land in long-tail search for the buyer-intent queries that actually convert.

### ERPClaw OS Phase 3

The self-extending architecture (the part that lets AI agents author new vertical modules within constitutional financial laws) is at Phase 2. Phase 3 ships in Q3 and adds the invariant engine to vertical generation, which means a generated module will have its GL behavior mathematically constrained at author time rather than at test time. There will be a separate launch post for this.

## Numbers, honestly

This is the part I commit to keeping in every quarterly post. No vanity metrics, no creative framing.

| Metric
| End of Q2 2026

| GitHub stars
| ~35 (across the org, eight weeks since first public commit)

| Site traffic (erpclaw.ai)
| TBD; site went live in current shape on 2026-04-26

| Total installs
| TBD; we do not phone home, so we count via opt-in registration only

| Paying customers
| 0 (intentional; cloud managed launches Q4)

| Revenue
| $0 (intentional)

| Back-office coverage
| Full suite: sales, purchasing, stock, manufacturing, HR, payroll, projects, and vertical packs (as of Q2 2026)

| API surface
| Deep, scriptable action set across the whole system (as of Q2 2026)

| Test depth
| L0 constitutional + L2 contract + L3 smoke + invariant engine, run on every relevant test (as of Q2 2026)

| Public blog posts
| 17

| Public docs pages
| 12

| Marketing pages total
| 73

A few notes on the zeros and TBDs.

Revenue is zero because cloud managed does not launch until Q4 and the self-hosted product is intentionally open source and free. The business model is hosted plus support contracts, not feature gates. This is a deliberate trade for a slower revenue ramp in exchange for a faster install curve and a defensible open source moat.

Install count is TBD because ERPClaw runs locally on the user’s machine and does not phone home. We can count the opt-in registration page submissions and the GitHub clone count, but neither is a clean install number. If we shift to phone home telemetry later it will be opt-in and disclosed; I am not sure we will.

Site traffic is TBD because the seventy-three page version of the site has been live for less than a week. The next quarterly post will have the first real curve to look at.

The number I am personally watching most closely is GitHub stars, because it is the cleanest leading indicator of community pull. Thirty-five stars in eight weeks is not a viral curve. It is the seed phase. The bet is that the comparison and vertical content the team is shipping in Q3 plus the Shopify App Store launch will move the slope. Q3’s quarterly post will tell whether that bet paid off.

## A note on the team

ERPClaw is built at AvanSaber Inc by a small co-founder team: Nikhil, Varun, and a handful of advisors. On the engineering side day-to-day, I pair with Claude Code as a coding companion, which is how a full marketing site, two deep integrations, a constitution-enforced test suite, and a full back-office module suite ship at this cadence. If you want the engineering writeup of how that workflow actually runs, building ERPClaw with Claude Code is the long form version.

The reason I am flagging this is not humble bragging. It is that the AI-assisted small team economic curve is the thing I am most certain about over the next five years, and ERPClaw is the artifact I am building to test that curve. Every quarterly post will keep the team count and the artifact count visible together so you can watch the ratio.

## How to follow along

If you want to actually try ERPClaw, the install is one command:

`clawhub install erpclaw`
Then talk to your assistant. It sets up your chart of accounts, suggests the modules you actually need, and gets you to a working ERP in about three minutes.

The docs are the right starting point if you want to read first. The demo is the right starting point if you want to see it move. The pricing page is the right starting point if you are trying to figure out what this is going to cost (still free; cloud managed pricing lands in Q4).

If you want to follow the build, the public repo is at github.com/avansaber. A star is the cheapest way to say the curve is bending the right way. An issue or a pull request is the most useful way.

The next quarterly post lands on the last Monday of July 2026. By then we will know whether the App Store listing landed, whether multi-currency v1 shipped on time, and whether the seventy-three page site is moving the traffic needle. I will tell you the truth either way.

That is what shipped. That is what is next. See you in Q3.

---

## ERPClaw + Stripe: Deep Coverage, ASC 606, $0
URL: https://www.erpclaw.ai/blog/erpclaw-stripe-deep-integration-launch/

# ERPClaw + Stripe: Deep Coverage, ASC 606, $0
ERPClaw's Stripe integration is in production: payout reconciliation in three layers, ASC 606 revenue recognition, and Connect fees. Free and open source.

Stripe is the easiest part of running a SaaS business. Stripe
accounting is the hardest. ERPClaw fixes the second part.

The Stripe integration in ERPClaw is now in production, and it is
free. Deep Stripe coverage including everything
you actually need: payments, payouts, fees, disputes, subscriptions,
ASC 606 revenue recognition, Connect platform
fees, three-layer payout reconciliation, MRR reports, and a clearing
account that hits zero or tells you why. No enterprise sales call,
no $50,000 minimum, no 45-day implementation.

If you run a SaaS company on Stripe and you have been quietly avoiding
the question of “wait, are these books actually right?”, read on.

## What “Stripe accounting” actually means

When a customer charges $100 on Stripe, the money does not arrive in
your bank for two to seven business days. Stripe holds the funds in a
clearing account. Stripe also takes a cut: typically 2.9% plus 30
cents per transaction for cards, more for ACH and international.
Stripe pays out periodically, in batches, and the batch totals do not
match individual charges.

Now layer on:

- Refunds that reduce a previous charge but not the previous fee

- Disputes that put funds in escrow until resolved

- Connect payments that route part of the money to a connected account

- Subscriptions that create predictable but compounding revenue

- Recognition rules that say you cannot count an annual subscription as revenue all at once

This is the surface that most accounting apps either give up on or
charge you a premium tier to handle properly.

## The three-layer reconciliation

ERPClaw reconciles in three layers, every one of which has to balance:

Layer 1: every Stripe transaction (charge, refund, fee, adjustment)
becomes a journal entry in the ERPClaw GL. The net of all journal
entries for a payout period equals the payout amount.

Layer 2: every Stripe payout becomes a deposit in your ERPClaw
bank account, matched to the actual bank statement. The deposit total
equals the bank credit.

Layer 3: the Stripe Clearing GL account in ERPClaw must net to
zero across the period. Charges and fees in, payouts out. If it does
not zero, the books are wrong and ERPClaw flags it.

This is the same pattern A2X uses for Shopify, except A2X pricing runs from $29 to around $229 a month depending on order volume (as of June 2026), and we do it for free across both Stripe and Shopify.

## ASC 606, the boring part that matters

If you are a SaaS founder, your auditor or future investor will care
about ASC 606. It is the accounting standard for revenue recognition,
and it says you cannot count a year of subscription revenue when the
customer pays it. You have to recognize it monthly over the
subscription period.

Most Stripe accounting tools give you a CSV export and tell you to
figure out ASC 606 yourself. ERPClaw posts the journal entries
correctly the first time:

- Customer pays $1,200 for an annual plan: ERPClaw posts $1,200 to
Cash and $1,200 to Deferred Revenue (a liability).

- Each month for twelve months: ERPClaw recognizes $100 by debiting
Deferred Revenue and crediting Subscription Revenue.

- Customer cancels mid-year: ERPClaw computes the unrecognized portion,
reverses it, and applies the refund correctly across all the right
accounts.

The actual rule is more nuanced (performance obligations, variable
consideration, contract modifications), but the engine is built on the
five-step ASC 606 framework and handles the standard SaaS cases
correctly. We document every rule we follow at
erpclaw.ai/docs/stripe so you
or your CPA can audit the logic.

## Connect platform fees, done correctly

If you run a marketplace or platform on Stripe Connect, you have an
extra layer of complexity: application fees, transfers to connected
accounts, and reverse transfers when refunds happen.

ERPClaw handles the standard Connect patterns:

- Direct charges with application fees: gross to the connected
account’s books, application fee posted as your platform revenue,
Stripe fee deducted from the connected account’s revenue.

- Destination charges: gross to your books, application fee held
back, transfer to the connected account.

- Separate charges and transfers: full visibility into both halves.

This was the most-requested feature from the founders we talked to
while building the integration. Most existing tools either ignore
Connect or charge enterprise pricing for it.

## MRR, ARR, and the metrics SaaS founders actually want

Stripe gives you raw transaction data. SaaS founders need:

- MRR: monthly recurring revenue, normalized to one month

- ARR: annualized run rate (MRR x 12)

- Net new MRR: new + expansion - contraction - churn

- Logo churn vs. revenue churn: distinct metrics, both matter

- Cohort retention: how does the cohort that signed up in March 2025 look today?

ERPClaw computes all of these from your actual Stripe subscription
data, not estimates. If a customer downgraded mid-cycle, the math
reflects that. If a customer paused then resumed, the math reflects
that. If your plan structure is complicated (per-seat, metered,
hybrid, multi-product), the math still reflects that because we
parse the Stripe subscription items directly.

## What this replaces

We are not subtle about who we are aiming at. ERPClaw replaces:

- Rillet, which charges enterprise pricing for individual-transaction
reconciliation that should be table stakes

- Synder, which is fine but priced from $65 to $275 a month (as of June 2026) and feature-limited

- A2X for Stripe, which exists but is mostly Shopify-focused

- Hand-rolled spreadsheets, which work right up until they don’t

If you are paying any of these, install ERPClaw on your own
infrastructure, point it at your Stripe account, and watch your books
reconcile in five minutes.

## What it does not do

ERPClaw is not Stripe Sigma. We do not run analytical SQL queries
against your Stripe data. (You can do that yourself with Sigma, or
export from ERPClaw.) ERPClaw is not a fraud-detection system; Stripe
Radar handles that. ERPClaw is not a billing system; Stripe Billing
handles that.

ERPClaw is the accounting and GL behind Stripe. The bookkeeping
plumbing that auditors, investors, and your future-self are going to
ask about.

## How to install

Two paths:

Stripe Marketplace (recommended, two clicks): the app is live at
marketplace.stripe.com/apps/erpclaw-accounting.
Install, follow the OAuth pairing flow, and you are live.

API key flow (power users, air-gapped installs): paste your Stripe
restricted key into ERPClaw via `stripe-add-account` and the integration
syncs without ever talking to the marketplace Worker.

Full install walkthrough at
erpclaw.ai/docs/stripe/install-walkthrough.

## Why we built this

Our co-founders spent years rolling out enterprise SAP, built top-rated operations software (ZapInventory was ranked third worldwide on Crozdesk in 2020), and have looked at every flavor of business software: enterprise, mid-market, SMB, hand-rolled.

The pattern is always the same: SaaS founders pay way too much for
accounting tools because the alternative is hiring a CPA firm or
running their books in a Google Sheet. The tools that exist are good,
but they are priced for venture-funded SaaS companies who can absorb a
$2K-per-month line item without thinking. The merchants who need them
most cannot afford them.

ERPClaw is what happens when you write the spec precisely, hand it to
Claude, and refuse to charge for commodity business logic.

## What is next

We are watching for any issues that turn up in the first week. The
deeper Stripe Marketplace listing is in submission. The full doc set
at erpclaw.ai/docs/stripe is
expanding by the week.

Star the GitHub repo if you want to follow along. Email
`support@avansaber.com` if you hit anything broken. And if you are
on Rillet, A2X, or Synder, install ERPClaw, run it in parallel for a
month, and tell me if the journal entries match. They should.

Time to install and start syncing.

---

## ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared
URL: https://www.erpclaw.ai/blog/erpnext-ai-ecosystem-2026-comparison/

# ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared
ERPNext core ships no native AI in 2026. The third-party Frappe app ecosystem filled the gap with NextAI, ChatNext, ChangAI, and an MCP bridge. Here is the map.

ERPNext by Frappe Technologies is one of the most-installed open-source ERPs on the planet. Roughly 100,000 production installs, fifteen years of community, a respectable global reseller network. When ERPNext users ask “how do I add AI to my install in 2026?”, the honest answer is that the core does not ship native AI and the ecosystem has filled the gap with a handful of third-party Frappe apps, each solving a different slice of the problem.

This post is a map. It walks through five of the most widely used AI options in the Frappe ecosystem in 2026, what each one does, what license each ships under, and where the tradeoffs are sharpest. At the end I will be honest about where ERPClaw sits in the same picture.

## ERPNext core in 2026 does not ship native AI

The architecture story explains why. ERPNext is built on the Frappe Framework, which is a forms-and-workflow toolkit. A DocType defines a form, a workflow defines the state transitions, and the UI is generated from the schema. MariaDB stores the data, Redis handles caching and the queue, Node.js renders the front end, Nginx fronts everything. It is a clean architecture, fifteen years polished, and it was designed for a world where the primary interaction model is a human filling out a form.

AI was not in scope when the architecture was set. There is no native chat copilot in ERPNext 15 (the current line in 2026). There is no agent runtime, no embedded LLM, no document-OCR pipeline shipped with the core. What the ecosystem has done instead is build add-on Frappe apps that you install separately into your ERPNext bench. Each app brings its own AI surface.

## NextAI

NextAI is one of the more widely installed AI apps in the Frappe ecosystem in 2026. The angle is content generation. NextAI adds AI-assisted writing across ERPNext records: drafting product descriptions, polishing email replies, generating customer-facing copy, fixing grammar, suggesting contextual improvements.

It plugs OpenAI as the default model provider. Some forks support Anthropic and local models via Ollama, but the main branch defaults to OpenAI and expects you to bring your own API key. The Frappe app is free, but the model calls are billable to the OpenAI account you wire in.

What NextAI is good at is taking the writing tax off founders and operations staff. What it does not do is operate ERPNext for you. You still log in to Desk, go to the right DocType, and use NextAI to improve the text in a single field. It is a writing assistant inside ERPNext, not an agent that runs ERPNext.

## ChatNext

ChatNext (also listed as “NextAI” on the Frappe Marketplace) is a sibling project from the same team and the naming is admittedly confusing. ChatNext is the conversational layer. The pitch is that a user can open a chat window and ask questions about ERPNext data in natural language: “what is the AR aging for Acme?”, “show open POs from last month over $5,000”, “which items are below reorder point?”.

Under the hood, ChatNext translates the question into a Frappe report query, runs it, and returns a formatted answer. It is essentially a natural-language query interface over the Frappe ORM, with an LLM doing the translation step.

This is more useful day to day than NextAI for an operations team because it cuts out the click-through-screens cost of finding data. It is still a query interface, though, not an action interface. ChatNext can read your data; it cannot, by default, create a sales invoice or post a journal entry for you. The latest builds have started adding write actions for a small set of common operations, but the coverage is narrow and uneven.

## ChangAI

ChangAI from ERPGulf is a broader AI integration for the Frappe stack. The scope is larger than NextAI or ChatNext on paper. ChangAI bundles a chat sidebar, document OCR for invoice and receipt processing, a forecasting helper for sales and inventory, and a hooks-based system that lets you wire AI into custom server scripts.

The breadth is appealing if you want one app instead of five. The tradeoff is that depth in each area is shallower than the focused single-purpose apps. ChangAI’s OCR is fine for clean invoices; for messy scans it lags the purpose-built OCR vendors. ChangAI’s chat is functional but the query coverage is narrower than ChatNext.

Licensing on ChangAI is GPL v3 like the Frappe core, which is helpful because it means there is no license-mixing concern when you install it next to ERPNext. Several of the other apps in this space use MIT or proprietary licenses, and the mix matters if you are publishing your own derived work.

## Aerele Chatbot

Aerele Chatbot takes a different angle. Instead of putting the AI interface inside the Frappe Desk, it puts ERPNext behind Telegram, WhatsApp, and Slack. You message a bot, the bot relays to ERPNext, and ERPNext does the work.

For sales teams that live in WhatsApp Business this is a real productivity win. The founder can type, for example, “invoice Globex $3,000 for May retainer” into Telegram and Aerele Chatbot will create the sales invoice, attach the PDF, and reply with a confirmation. That is a real action-layer use case, and Aerele is one of the few Frappe apps that actually closes the loop on actions rather than just reads.

The limits show up in coverage. Aerele Chatbot supports a curated list of actions out of the box; anything outside that list requires custom Frappe code and a Frappe developer who knows the chatbot’s request format. The integration is also model-agnostic but defaults to OpenAI, with the same bring-your-own-API-key story as NextAI.

## MCP integration via Composio

The newest entry in 2026 is the Composio ERPNext toolkit, which exposes ERPNext as a Model Context Protocol (MCP) server. The implication is that any MCP-aware AI client (Claude Desktop, Cursor, the new wave of agentic frameworks) can connect to ERPNext and call its operations as tools.

This is structurally different from the other four apps. NextAI and ChatNext live inside ERPNext and bring their own UI. Composio’s MCP bridge inverts the relationship: ERPNext becomes a service that external AI agents talk to. If you are an ERPNext user who wants to use Claude or another agentic platform as your primary interface and have it operate ERPNext as one of many tools, Composio is the right path.

The trade is that you are now running an MCP bridge service alongside ERPNext, configuring auth between the bridge and your ERPNext instance, and depending on Composio’s hosted infrastructure (or self-hosting their stack). It is more moving parts than the in-Desk apps.

## The tradeoffs of the third-party-app model

What this map shows, when you step back from it, is the cost of being AI-decorated instead of AI-native. To get a reasonably full AI footprint on ERPNext in 2026 you typically install:

- NextAI for content generation.

- ChatNext for natural-language queries.

- ChangAI or a similar app for document OCR.

- Aerele Chatbot for messaging-channel actions.

- Composio MCP for agent platform integration.

Five separate apps, each with its own:

- Install and upgrade path. Each app version-pins independently of ERPNext. When ERPNext 16 ships, you will wait for each app’s maintainer to catch up.

- License. GPL v3, MIT, AGPL, and proprietary all appear in the mix. License compatibility matters if you are deploying internally with strict policy or if you are reselling.

- Model provider. Each app picks its default LLM. You will end up with API keys for OpenAI, Anthropic, and possibly Google in the same install, each billed separately.

- Support story. Frappe support is for ERPNext; each app’s support is for that app. When something breaks at the boundary, the back-and-forth between vendors is yours to coordinate.

- Data flow. Each app sends a different subset of your ERPNext data to a different external model. Mapping out who-sees-what is its own audit.

This is not a knock on the Frappe community. It is what an AI-decorated architecture forces. Because the AI surface is bolted on, the integration overhead is real and falls on the user. For an established ERPNext shop with a Frappe consultant on retainer, this is manageable. For a founder doing it themselves, it is a lot.

## ERPClaw as a different shape of answer

ERPClaw is built the other way. Honest disclosure, I am the builder. AI is the primary interface from the first commit, not a sidebar added later. That changes a few things in the install footprint:

- One install, one license. GPL v3, free forever, self-hosted on your own infrastructure. No per-user fee, no Enterprise tier, no add-on apps to license separately.

- The action layer is the AI’s interface. Every operation an AI agent might want to take, create a customer, post a journal entry, submit an invoice, reconcile a Stripe payout, is exposed as a versioned action with a typed contract. The same actions power the CLI, the chat clients, the MCP bridge, and the web UI.

- Audit-grade accounting under the hood. Money is stored as exact decimals (not floats). Every GL posting passes a 12-step validation. Submitted entries are immutable; corrections are reversals, not edits.

- Cryptographic verification of every release. Each install resolves against a signed module manifest, so the code you got is provably the code we shipped.

- 5-minute install. One command pulls the foundation and the verticals you need. Auto-detection picks the right industry pack.

The honest comparison against ERPNext, including the places where ERPNext is genuinely ahead (community size, localization breadth, certified-consultant network), lives in ERPNext vs ERPClaw: An Honest AI-Native Open-Source ERP Comparison. The head-to-head pricing and feature snapshot is at /compare/erpnext/.

Two related reads on competitor AI stories in 2026:

- SAP Joule and SAP Business One: What’s Embedded vs What Isn’t

- Odoo’s Enterprise AI Pricing in 2026: What $24.90/user/month Actually Buys

## The summary

ERPNext in 2026 is a strong open-source ERP with no native AI in the core and a healthy third-party Frappe app ecosystem that has filled the gap. NextAI handles content generation. ChatNext handles natural-language queries. ChangAI bundles a broader feature set. Aerele Chatbot bridges messaging channels. Composio MCP exposes ERPNext to agentic AI platforms. Together they cover most of what an AI surface for an ERP looks like in 2026, at the cost of running five apps with five maintainers and five licenses.

If that integration tax fits your team’s profile, ERPNext plus the ecosystem is a workable AI-decorated path. If you would rather start from a stack that put AI in the core from line one, ERPClaw is one of a small group of alternatives worth a serious look.

Compare ERPClaw and ERPNext side by side.

Tagserpnextfrappeai-erpcomparisonopen-source-erp
Share:XLinkedInHacker News

---

## ERPNext Alternatives 2026: the AI-Native Take
URL: https://www.erpclaw.ai/blog/erpnext-alternatives-2026/

# ERPNext Alternatives 2026: the AI-Native Take
A vendor-neutral guide to the real ERPNext alternatives in 2026, including the open-source options that are also AI-native, not just AI bolted on via plugins.

Short answer. Unlike most “alternatives” searches, this one is rarely about price: ERPNext is already free and open source. People look for an alternative for three reasons. They find ERPNext heavy to set up and maintain (look at simpler open-source options, or a managed product). They want AI that does the work, not a plugin bolted on (look at the AI-native tools, and at ERPClaw). Or they want to leave self-hosting entirely (a paid cloud ERP). This guide covers all three honestly, and gives you a test to separate genuine AI-native from AI added on.

ERPNext earned its place. The Frappe team has shipped for many years, roughly 100,000 installs run real businesses, and the DocType model is genuinely clever. If a consultant configured it for your company last year and your ops team is trained on it, this post is not asking you to switch tomorrow. It is for the person who looked at the setup, the maintenance, or the AI story and started asking what else exists now.

## The test: genuine AI-native, or AI bolted on?

ERPNext is open source, which is excellent. On AI, though, it is in the same boat as the paid incumbents: AI arrives through plugins and add-ons, not the core design. Before you shortlist anything, run three questions:

- Does the AI do the work, or describe it? A model that summarizes a report is not an agent that posts the journals and reconciles the accounts. Ask to see an action completed end to end.

- Was the system built around AI, or retrofitted? Bolting a model onto an existing codebase keeps the old workflows. Software specced around AI from the start behaves differently.

- Can you talk to it to get something done? “Raise a purchase order for 200 units from Acme and receive it against the open requisition” should just happen. If the only interface is forms with a chat sidebar, that is AI-decorated, not AI-native. Full argument in AI-decorated vs AI-native software.

Hold each option below to that test.

## Group 1: other open-source ERPs

If you like owning your software but want a different fit than ERPNext, these stay in the open-source camp.

- Odoo. The other big open-source ERP. Broad, modular, large community. Like ERPNext, AI comes via apps rather than the core, so it is open source but AI-decorated. See the Odoo alternative take.

- Dolibarr and similar. Lighter open-source options for smaller teams; capable for basic ERP and CRM, thinner on deep accounting.

- ERPClaw. This is the one we build, so weigh it accordingly. The difference from ERPNext is the one that matters here: it is open source and AI-native. The agent posts double-entry journals end to end rather than summarizing a screen, and the install is minutes rather than a setup project. Same $0, self-hosted, you own the data. It covers the back-office primitives (accounting, inventory, manufacturing, HR, payroll, CRM, projects, billing) in one shared database. The honest gaps versus ERPNext: a smaller community and fewer pre-built vertical apps today. We are newer.

## Group 2: the AI-native tools

If the real reason you are leaving ERPNext is that the AI is bolted on, this is the category built the other way.

- Rillet and DualEntry. AI-native finance platforms designed around automation from the start. Genuinely pass the test above. The trade-off: closed source and paid, so you are back to renting.

- ERPClaw. The open-source member of this group, which is the point: you get the AI-native architecture without giving up ownership or paying per seat.

## Group 3: leave self-hosting entirely

If the maintenance burden is the real issue and you would rather someone else run it, a managed cloud ERP is the move. These are paid and AI-decorated, but you stop patching servers.

- NetSuite, Sage Intacct, Acumatica, Dynamics 365 Business Central. Established managed ERPs at a range of price points. See our NetSuite alternatives and Sage Intacct alternatives guides if you go this direction.

## Side by side

| Option
| AI
| Open source
| Self-host
| Setup
| Price posture

| ERPNext
| Plugin
| Yes
| Yes
| Project
| Free or paid cloud

| Odoo
| Plugin
| Yes
| Yes
| Project
| Free or paid tiers

| Dolibarr
| Plugin
| Yes
| Yes
| Moderate
| Free

| Rillet / DualEntry
| Native
| No
| No
| Onboarded
| Paid

| NetSuite / Sage Intacct
| Add-on
| No
| No
| Months
| Paid

| ERPClaw
| Native
| Yes
| Yes
| Minutes
| $0 forever

The table is not about ERPClaw winning every row. It is that open source, AI-native, and minutes-to-install in the same row is a near-empty cell. ERPNext gives you the first; ERPClaw gives you all three.

## Which group are you in?

- You want open source but a different fit. Odoo for breadth, ERPClaw if the missing piece is genuine AI plus a fast install.

- You want AI to actually run the work. Rillet or DualEntry if you will pay and go closed; ERPClaw if you want that architecture and keep ownership.

- You are done self-hosting. A managed cloud ERP. Just know you are trading ownership for someone else running it.

## FAQ

Why look for an ERPNext alternative if it is already free?

Cost is rarely the reason. The usual ones are setup and maintenance effort, and the fact that AI in ERPNext is added through plugins rather than built into the core. If those do not bother you, ERPNext is a strong choice and worth keeping.

Is there an open-source ERP that is actually AI-native?

Yes. ERPClaw is open source and AI-native, where the agent completes accounting actions end to end rather than summarizing them. ERPNext and Odoo are open source but reach AI through add-ons.

How hard is it to move off ERPNext?

It is a real project, mostly because the DocType customizations and data have to be re-mapped. Run the new system in parallel through a period close, validate the trial balance to the cent, then cut over.

How does ERPClaw compare to ERPNext directly?

See ERPClaw vs ERPNext for the side by side and the honest founder comparison for the detailed tradeoffs.

## Where to go next

- ERPClaw vs ERPNext, the detailed side by side

- ERPNext vs ERPClaw: an honest comparison

- Odoo alternative: an AI-native take

- The 5 AI-native ERPs that actually earn the label

If you want an open-source option that is also AI-native and installs in minutes, try ERPClaw alongside your current setup for a period. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

Tagserpnextalternativesai-nativeopen-sourceerpcomparison
Share:XLinkedInHacker News

---

## ERPNext vs ERPClaw: an Honest Open-Source Comparison
URL: https://www.erpclaw.ai/blog/erpnext-vs-erpclaw-honest-comparison/

# ERPNext vs ERPClaw: an Honest Open-Source Comparison
ERPNext is a proven open-source ERP with a 15-year track record. ERPClaw is AI-native, modular, open source. Honest tradeoffs from a founder who has shipped both styles.

ERPNext is a serious piece of software. I want to say that up front, because most “ERPNext alternative” posts open with a list of grievances and that is not where I am coming from. The Frappe team has been shipping for fifteen years. There are roughly 100,000 ERPNext installs running real businesses. The DocType system is genuinely clever. If you spent six months last year configuring ERPNext for your company and a Frappe consultant trained your ops team, this post is probably not asking you to switch tomorrow.

This post is for a different person.

This post is for the founder who looked at ERPNext nine months ago, deployed it on a 4 GB VPS, got it running, and is now adding a ChatGPT plugin so the sales team can ask “what is the AR aging for Acme” without clicking through six DocType screens. Or the ERPNext admin who keeps a separate Stripe-to-ERPNext sync script running on a cron job because the built-in payment integration does not understand application fees. Or the SMB owner who is comparing two open-source options for the first time and wants to know if the AI-native angle is marketing or architecture.

ERPClaw is the alternative. Free, open source, AI-native by design, full operations surface across 14 verticals. Single SQLite file by default, or PostgreSQL if you prefer (we use PyPika so the same code targets either). The claim is not “we beat ERPNext at being ERPNext.” The claim is that the architecture underneath is structurally different, and the difference matters most for the workflows ERPNext was never built around: chat-first input, deep payment-rail integrations like Stripe and Shopify, and a system that the AI itself can extend.

ERPNext is open source like us, but it is monolithic, form-driven, and AI-decorated rather than AI-native. The full argument on that distinction lives in AI-decorated vs AI-native software. The short version: bolting a chat sidebar onto a forms application is not the same thing as designing the system so the AI is the primary surface.

## What ERPNext does well

I am not interested in trashing a product that 100,000 businesses run on. There are real strengths here.

The community is enormous. Fifteen years of forum threads, a public conferences circuit, a network of certified consultants in India, the US, and Europe, and a third-party app marketplace. When you Google an ERPNext question, you get answers. ERPClaw is two months old in public; that depth does not exist yet.

Frappe is a real framework. The DocType system, the report builder, the print format designer, the workflow engine, and the role permission model are parts of one coherent abstraction. If you know Frappe, you can build a custom CRM module in a weekend.

The track record is real. Manufacturing companies with hundreds of users run ERPNext in production. The accounting module passes statutory audits in dozens of countries. Localization coverage (India GST, EU VAT, Saudi ZATCA, and more) is wider than ERPClaw’s today.

The customization story is mature. Custom DocTypes, server scripts, client scripts, hooks, and the Frappe Builder give you many ways to extend without forking. If you have already paid the learning cost, those advantages compound.

## Where ERPNext stops

The strengths above describe a 2010 vision of what an ERP should be: a forms application on top of a strong framework, hosted on a real server, customized by trained developers. That vision is not wrong. But four things have changed since then, and ERPNext was not designed for any of them.

Form-driven UX is friction in 2026. The default ERPNext interaction is: log in to Desk, open the right DocType list view, click New, fill in twelve fields, save, submit. For a trained user that is fine. For example, a founder who wants to type “invoice Acme $5K for April consulting, due in 30 days” into Telegram hits a wall. ERPNext has chat plugins, but the application underneath is still forms; the chat is a translation layer over the forms layer rather than the native interface.

The footprint is heavy. A standard ERPNext install needs MariaDB or PostgreSQL, Redis (often three instances for cache, queue, and socketio), Node.js for the asset pipeline, nginx, supervisord, and the bench tool. Minimum recommended RAM is 4 GB and most production installs sit on 8 GB or more. The first-time install on a fresh Ubuntu box is hours of “why did this step fail” if you have not done it before.

Monolith over modules. ERPNext is one application. You install it and you get accounting, selling, buying, stock, manufacturing, projects, HR, payroll, support, and the website builder, all in one bench. You cannot install only accounting. Custom Apps give you composition, but the base monolith is always there.

AI is bolted on, if at all. Frappe has experimented with AI features (smart suggestions, document summarization), but they are sidebar features over a forms-based product. There is no native concept of “the AI calls actions on your behalf and posts to the ledger inside a single transaction.” If you want that on ERPNext today, you write a custom server script, expose it via REST, and wire up an OpenAI or Claude integration yourself. People do it. It is not the default.

No native deep integrations for modern payment rails. ERPNext has a Stripe payment gateway plugin (it generates a checkout link). It does not have a deep Stripe sync that pulls every event, posts to the GL with a clearing-account pattern, recognizes ASC 606 revenue, and handles Connect platform fees with three-layer reconciliation. It does not have an equivalent Shopify integration for orders, refunds, gift cards, and payouts. ERPClaw ships both in the core.

These are not defects. They are choices. ERPNext is what it is on purpose, and for a lot of buyers in 2018 it was the right choice. In 2026 the choices land differently.

## What AI-native architecture means in practice

This phrase gets thrown around. Most products that say “AI-native” mean they shipped a chat sidebar last quarter. So I want to be specific about what it means for ERPClaw.

The spec is the source of truth. ERPClaw is built spec-first: a markdown SKILL.md per module describes the actions, inputs, outputs, GL posting rules, and validation. The Python code is generated and regenerated from those specs by Claude Code. When a new vertical needs a new module, ERPClaw OS generates it from proven module patterns and validates the result against the constitutional rule set before it ships. That is a different production model from “developers write Python forever.”

The UI is chat-first, not forms-with-chat. The primary interface for ERPClaw is OpenClaw, an AI runtime on Telegram, WhatsApp, the web dashboard, or any client that speaks to a Claude or GPT backend. You type “show me AR over 30 days for Acme.” The AI calls the right ERPClaw action, gets back JSON, and answers in plain English. There is no Desk to log into.

Modules are skills, not customizations. ERPNext extensions are Custom Apps inside the Frappe framework. ERPClaw extensions are independent modules that publish their own actions, own their own tables, and live in their own GitHub repos. The module catalog is a registry of skills the AI can install on demand. When a user says “I run a dental practice,” ERPClaw pulls healthclaw-dental from GitHub, runs init_db.py, and the AI immediately understands dental workflows. No “rebuild the assets, restart the workers, migrate the database” cycle.

Database is your choice. ERPClaw defaults to a single SQLite file, which is enough for the vast majority of SMBs (WAL mode, FK enforcement, tens of millions of rows). For larger installs or teams that already run a Postgres cluster, the same code targets PostgreSQL through the PyPika query builder. We are not SQLite-only. You pick the database your situation needs, not the one the framework demands.

These four points together are what AI-native means here. Each one shows up in the install footprint, the daily UX, and the cost of adding new functionality.

## Side by side

The full comparison is on /compare/erpnext. Pulling the table here for reference:

| Aspect
| ERPClaw
| ERPNext

| Price
| Free (open source license)
| Free (GPL) or $50+/mo cloud

| Interface
| Natural language chat
| Traditional web UI (Desk)

| Setup time
| 5 minutes
| Hours to days

| Training
| None needed
| Significant learning curve

| Infrastructure
| SQLite single file or PostgreSQL via PyPika
| MariaDB plus Redis plus Node.js plus nginx

| Server requirements
| 2 GB RAM, any Linux or macOS
| 4 GB RAM minimum, specific stack

| Database backup
| Copy a file (or `pg_dump`)
| `mysqldump` plus Redis backup

| AI integration
| AI-native (every interaction)
| Bolt-on (limited)

| Framework dependency
| None (pure Python plus your DB choice)
| Frappe framework required

| Deployment
| Copy files to server
| bench setup plus bench install

| Customization
| Edit Python scripts directly
| DocType system plus hooks

| Multi-tenant
| Multi-company in single DB
| Separate sites per tenant

| License
| GPL v3 (copyleft)
| GPL (copyleft restrictions)

| Stripe integration
| Deep event sync, ASC 606, Connect
| Basic gateway plugin

| Shopify integration
| Deep order sync, GDPR webhooks
| Third-party app required

Two rows deserve a second look.

Infrastructure. ERPNext on a fresh Ubuntu box is roughly an afternoon of bench install, MariaDB tuning, Redis configuration, supervisord setup, and asset compilation. ERPClaw on the same box is `clawhub install erpclaw` and you are done in five minutes. If you want PostgreSQL instead of SQLite, you set one environment variable. The reason this matters is not that “setup is hard for ERPNext” (people do it every day), it is that the operational burden continues forever. Backups, upgrades, log rotation, and security patches all scale with the number of moving parts. ERPClaw has fewer moving parts on purpose.

License. Both ERPClaw and ERPNext are GPL v3 (effective ERPClaw v4.2.0; v4.1.x and earlier were MIT). License is a checkbox; the real wedge is architecture. ERPClaw is AI-native by design (the action layer is the API; every operation invokable from a prompt with full audit trail). ERPNext is forms-and-workflows with AI plug-ins (changAI, NextAI, Ollama integrations) layered on. Both are open, both are copyleft, both can be self-hosted. Pick by architecture, not license.

## When to stay on ERPNext

I promised honest. Here are the cases where I would tell you to stay on ERPNext and not even bother with ERPClaw.

You have already paid the customization tax and your team is trained. If you spent four months last year configuring custom DocTypes, writing server scripts, building custom print formats, and training your ops team on the Frappe Desk, that investment is real. Throwing it out to chase a newer architecture is a bad trade. Stay on ERPNext, keep iterating, and revisit in a year.

Your business runs in a country where ERPClaw localization is thin. ERPClaw covers US, Canada, UK, EU, and India today. ERPNext covers a wider set, including Saudi Arabia (ZATCA), UAE VAT, Brazil NFe, and others. If you need a country-specific tax form ERPClaw does not yet ship, ERPNext is the safer pick.

You need the public marketplace of third-party apps. ERPNext has a real ecosystem of paid and free apps for things like advanced HR, education modules, hospital management, and more. ERPClaw has a 46-module first-party catalog that covers a lot, but the third-party app surface is essentially zero today. If a specific ERPNext app is the reason you are evaluating, that is a legitimate reason to stay.

Your CTO insists on a Postgres-or-MariaDB-only stack and views SQLite as a toy. ERPClaw runs on PostgreSQL too, but if the cultural objection is “anything that can run on SQLite is not enterprise,” that is not a battle worth fighting in a comparison post. Pick the tool the team will respect.

If none of the above apply, ERPClaw is genuinely worth a weekend.

## How a migration actually works

The full step-by-step is going to live at `/blog/migrating-from-erpnext-to-erpclaw` (publishing Q1 2027). Short version so you can scope it.

Step 1. Export from ERPNext. Run `bench export-doc` for master data (customers, items, suppliers, employees, accounts) and use the report builder to export trial balance, AR, AP, and stock balances as of cutover. About an hour.

Step 2. Map your chart of accounts. ERPClaw ships a ready-made US chart of accounts. ERPNext charts are usually larger. The mapping script reads the export and produces a one-to-one mapping with manual review for custom accounts. One to three hours.

Step 3. Pick a cutover date. Start of a fiscal month. Close ERPNext through the prior month, lock the period, open ERPClaw with opening balances as of day one of the new month.

Step 4. Import opening balances. One command loads the trial balance as opening entries. The 12-step GL validation engine rejects any imbalance, so you know the import is clean.

Step 5. Import master data. Customers, suppliers, items, employees, BOMs. Each is a one-line import. About an hour.

Step 6. Run parallel for one month. Post the same transactions to both, compare trial balances at month-end. They should match to the cent. If they do, retire ERPNext. If they do not, find the discrepancy first.

For a 25-person company with moderate customization, the whole migration is two to four weekends.

## FAQ

Will my Frappe consultant still be able to help me?

Probably not directly. ERPClaw is pure Python plus your database; there is no Frappe framework, no DocType system, no bench. A Frappe consultant could learn ERPClaw (it is simpler), but the Frappe-specific skills do not transfer. If your consultant relationship is the main reason you are on ERPNext, factor that in.

Does ERPClaw support custom DocTypes?

ERPClaw has actions and tables instead. The equivalent of a custom DocType is a custom module: a folder with a SKILL.md, an init_db.py, and a db_query.py. Building one is mostly writing the spec and letting the AI generate the code. Faster than building a DocType by hand, but a different mental model.

What about multi-currency and consolidation?

In the core, not a paid tier. Transact in any ISO currency, hold balances in multiple functional currencies, consolidate in your reporting currency. ERPNext supports the same; this is parity.

Is the audit trail good enough?

Yes. The ledger is locked: entries cannot be edited after they are posted, and a cancellation creates a reverse entry on top. Every cross-table write happens in a single database transaction. Every action passes the 12-step validation pipeline. ERPNext has a similar story; parity here too.

What if ERPClaw disappears?

You still have the software. open source license, your code, your data, your server. The repo at github.com/avansaber/erpclaw is forkable by anyone.

Can I run both side by side during evaluation?

Yes. ERPClaw runs in a single SQLite file at `~/.openclaw/erpclaw/data.sqlite` and does not touch your ERPNext install. Bring up an ERPClaw test instance on a low-cost server, post a week of transactions to both, compare the outputs.

## Try it

Install takes five minutes. The whole ERP, covering the full back office, on your machine.

`clawhub install erpclaw`
That puts the core ERP (full accounting, inventory, HR, payroll) on your server. From there, talk to it: “I run a 12-person manufacturing shop in Ohio, set me up.” It picks the industry vertical, installs the relevant modules, generates the chart of accounts, and is ready for transactions.

If you want to read first:

- /features for the full module breakdown

- /pricing (it is free)

- /compare/erpnext for the side by side

- /features/erpclaw-os for how the AI-native architecture works under the hood

- /docs for install and the ERPNext migration guide

- /blog/sqlite-or-postgresql-database-choice for the database story

- /blog/ai-decorated-vs-ai-native-software for the AI-native architecture argument

If you want to read the ERPNext side, the official sites are erpnext.com and frappeframework.com. The team there has earned a long, honest look. We are not in the business of telling you they are wrong.

Repo: github.com/avansaber/erpclaw. Star it if you want to follow along, we ship weekly. Email `support@avansaber.com` if you hit something broken.

ERPNext is the right software when you want a proven, customizable, traditional ERP and your team is ready to invest in Frappe. ERPClaw is the right software when you want an AI-native ERP with a light footprint, a chat-first interface, and deep integrations for the payment rails most SMBs actually use. The line between those two is not about which team is smarter; it is about which architecture matches the way you want to work in 2026.

Time to install.

---

## Free AI Accounting Tools 2026: What Free Buys You
URL: https://www.erpclaw.ai/blog/free-ai-accounting-tools-2026/

# Free AI Accounting Tools 2026: What Free Buys You
Free AI accounting tools in 2026: Wave, Zoho Books, Puzzle, Odoo, GnuCash and ERPClaw compared, including what each free tier covers and where the paywalls sit.

Short answer. “Free AI accounting tools” covers three very different offers in 2026, and it pays to know which one you are looking at before you sign up. Free tiers of paid SaaS (Wave, Zoho Books, Puzzle) give you a real ledger at $0 but cap the business size or meter the AI. Free open-source ledgers (GnuCash, Odoo Community) are free forever but ship little or no AI. And one option is open source, $0, and AI-native at the same time: ERPClaw, the one we build, so weigh our take accordingly. This roundup goes tool by tool, with pricing checked against each vendor’s page as of July 2026, and states where each paywall sits before you hit it.

The search “free AI accounting tools” (and its sibling, “AI accounting software free”) has become one of the most gamed queries in business software. Every vendor with a trial and a chat sidebar shows up. So two ground rules for this list. “Free” means a tier you can run indefinitely at $0, not a trial. “AI” means the product does accounting work for you (categorization, matching, extraction, posting, reconciliation), not that a chatbot answers questions about the help docs.

## The three shapes of free

- The free tier of a paid product. Wave, Zoho Books, and Puzzle run this model. The ledger is free; the growth path is not. Nothing dishonest about it, but the tier is priced marketing: the vendor is betting your business outgrows the cap.

- Free open source, AI optional. GnuCash and Odoo Community cost nothing and never will, but the AI is either absent or lives on the vendor’s paid meter.

- Free, open source, and AI-native. The near-empty cell of the matrix. ERPClaw is the tool we found there, and since we build it, run our claims through the same test you run everyone else’s.

Nikhil Jathar, who co-founded ERPClaw at AvanSaber, puts the pattern this way:

“In accounting software, free usually means the ledger is free and the intelligence is rented. Read the pricing page from the bottom up. The AI line is where the meter is.”

Six tools cleared the bar for this roundup. Here is each one, with the catch stated plainly.

## Wave: free ledger, paid automation

Wave’s Starter plan is free and genuinely useful: unlimited invoices, estimates, bills, and bookkeeping records, with no time limit. For a service business that writes ten invoices a month, it is a fine place to keep books, which is why it anchors this list.

The catch is where the automation sits. On the free plan you enter transactions by hand. Automatic bank transaction import and auto-categorization, the features doing the AI-adjacent work, require the Pro plan at $19 a month or $190 a year, and receipt scanning is a separate add-on at $11 a month on the free plan ($8 on Pro), all as of July 2026. Wave also earns on payments: card processing runs 2.9% plus $0.60 per transaction. None of this is hidden. But the “AI” part of “free AI accounting” is precisely the part that is not free.

## Zoho Books: strong free tier, hard revenue cap

Zoho Books’ free plan is the most complete hosted free tier on this list, with one condition attached: it is available only while your business revenue for the financial year stays under $50K (as of July 2026). Inside the cap you get one user plus one accountant seat, up to 1,000 invoices a year, bank reconciliation, a customer portal, and the AI-adjacent piece, 50 receipt autoscans a month for document extraction.

The catch is the cap itself. Cross the revenue line and the free plan is no longer available; paid plans start at $20 a month, or $15 billed annually. The autoscan allowance is real but narrow, and support on the free plan is email only. It is a well-built tier for a very small business, designed to stop being available exactly when the books get interesting.

## Puzzle: the most AI-forward free tier, on a lifetime meter

Puzzle is the most AI-shaped product with a free tier that we found: built for startups, cash and accrual books, auto-drafted financial statements, and aggressive auto-categorization. The Starter plan is free until $20K of transaction volume and includes 25 AI credits (as of July 2026).

Read the credit line carefully: those 25 Starter credits are a lifetime allowance, per Puzzle’s pricing page. The recurring AI supply starts on the Complete plan at $50 a month billed annually, which carries 100 AI credits a month plus the AI accuracy review and reconciliation features. Puzzle is upfront about its meter, and the product underneath is genuinely built around AI. But the free tier is a demonstration of the AI, not a supply of it. Closed source, cloud only.

## Odoo: a free app, a free edition, AI on credits

Odoo is really two “free” offers. One App Free gives you a single Odoo app with unlimited users, hosted by Odoo, at no cost. Odoo Community is the open-source edition (LGPLv3), self-hosted and free forever, with real double-entry accounting inside a broad ERP.

The AI catch has two layers. The document-digitization AI (scanning vendor bills and expenses) runs on in-app purchase credits that are not included in any subscription, per Odoo’s own pricing page. And the wider feature set most buyers actually want sits in the paid Enterprise plans, which run around $31.10 per user per month for the US Standard tier billed annually (as of July 2026). We took the AI pricing apart line by line in the Odoo Enterprise AI pricing breakdown. Odoo Community is genuinely free software; free AI it is not.

## GnuCash: genuinely free, genuinely no AI

GnuCash is the control group. Free, GPL-licensed desktop double-entry accounting, maintained since the late 1990s, no cloud dependency, no upsell anywhere in the product. If the requirement is “free accounting software, full stop,” it remains one of the most honest answers available.

It also has no AI at all. No categorization model, no matching engine, no assistant. People do wire scripts and language models around it, but that is a project you build and maintain yourself, not a feature that ships. GnuCash earns its place here as the baseline: this is what free looks like when nobody is trying to convert you to a paid plan.

## ERPClaw: $0, open source, AI-native (ours)

ERPClaw is the tool we build, so read this section most skeptically. It is a full ERP with accounting at the center (an immutable double-entry general ledger with validation enforced at write time, plus inventory, billing, and the rest of the back office sharing one database), licensed GPL v3, self-hosted, and $0 forever. Not a free tier: there is no paid tier for the self-hosted product. The pricing page is short on purpose.

The AI part is architectural rather than a checkbox. The agent is the primary interface: you describe what happened, and it posts the journal entries end to end, with the ledger invariants checked before anything touches the books. That is the line we draw between AI-native and AI-decorated, and the fuller argument lives on the AI bookkeeping pillar and in the open-source AI accounting roundup. On the bank side there is a Plaid connector for linking bank accounts, plus bank statement import (OFX, CAMT.053, MT940, BAI2) with an auto-match reconciliation engine for statement-driven workflows. The Stripe and Shopify integrations post into the same ledger, and none of it is metered, because there is no meter.

The honest catches: you host it yourself (a laptop or a small server is enough), the primary interface is an AI agent in a terminal rather than a polished web dashboard, and the ecosystem is younger than Wave’s or Zoho’s. If “free” must also mean “hosted for me,” ERPClaw is not that today.

## Side by side, as of July 2026

| Tool
| Free forever?
| The cap
| AI in the free tier
| Open source
| Self-host

| Wave Starter
| Yes
| Manual entry; automation is Pro
| None (receipt OCR is a paid add-on)
| No
| No

| Zoho Books Free
| While revenue is under $50K/yr
| 1 user, 1,000 invoices/yr
| 50 receipt autoscans/mo
| No
| No

| Puzzle Starter
| Until $20K transaction volume
| 1 user
| 25 AI credits, lifetime
| No
| No

| Odoo One App Free / Community
| Yes
| One app, or DIY hosting
| Digitization billed via credits
| Community: yes (LGPLv3)
| Community: yes

| GnuCash
| Yes
| None
| None
| Yes (GPL)
| Yes (desktop)

| ERPClaw
| Yes
| None
| All of it
| Yes (GPL v3)
| Yes

Prices and caps per the vendor pages linked above, as of July 2026. Check them before you decide; free-tier terms move.

## The three paywalls “free” hides

Across the whole list, the same three patterns recur:

- The usage cap. Zoho’s revenue line, Puzzle’s volume line, Wave’s manual-entry ceiling. The tier is free until the business is real.

- The AI meter. Autoscans per month, lifetime credits, digitization credits. The ledger is free; the intelligence is metered separately.

- The exit toll. Closed-source free tiers keep your books in the vendor’s cloud. Leaving means a CSV export and a re-import, and the fidelity of your history depends on what the export includes.

None of these make a tool bad. They make “free” a term worth auditing. Jathar again:

“My test for free software is to stop paying and see what you still own. On most free tiers the answer is a CSV export. With open source the answer is everything: the code, the data, and the AI that runs the books.”

## Which one fits you

- Service business, a handful of invoices a month, zero appetite for setup. Wave Starter. Accept manual entry, or pay for Pro when the typing gets old.

- Very small product business under the revenue cap. Zoho Books Free is the most complete hosted free tier going.

- Funded startup with a bookkeeper or CPA watching the books. Puzzle’s Starter tier is a good on-ramp; budget for Complete once the lifetime credits run out.

- You want free software forever and do not care about AI. GnuCash, or Odoo Community if you want ERP breadth and will host it.

- You want the AI to do the bookkeeping, at $0, and you will self-host. ERPClaw. That empty cell of the matrix is the reason it exists. If your books revolve around a Shopify store, start with the free Shopify accounting guide; if you are weighing it against a paid ledger you already run, the open-source Xero alternative walkthrough covers the switch.

## FAQ

Is there a genuinely free AI accounting tool in 2026?

Yes, with a distinction. Free tiers that include some AI exist (Zoho’s monthly autoscans, Puzzle’s credits), but the AI is metered and the tiers are capped. Unmetered free AI accounting means open source and AI-native together, and ERPClaw is the option we know of in that cell. If you know another, file an issue on the ERPClaw repo and we will update this roundup.

What is the best free AI accounting software for a small business?

It depends on which cap you hit first. If you invoice a few clients and want hosted simplicity, Wave. If you are under Zoho’s revenue cap and want the fullest hosted tier, Zoho Books. If you are a startup with investors, Puzzle. If you can self-host and want no caps at all, ERPClaw.

What does the phrase “AI accounting software free” usually get wrong?

It conflates free trials with free tiers, and free ledgers with free AI. Most results are trials. Of the true free tiers, the AI is usually the metered part, so check the credit and scan allowances before assuming the automation is included.

Can I run real books on a free tier without lock-in?

Only if you can take the full ledger with you. Hosted free tiers hold the books in the vendor’s cloud, so your exit is whatever their export produces. Self-hosted open source (GnuCash, Odoo Community, ERPClaw) keeps the ledger on your disk, which is what “your data” means in practice. The longer argument is on the open-source AI accounting pillar.

## Where to go next

- AI bookkeeping: what AI-native means in the bookkeeping cut

- Open Source AI Accounting: the 2026 honest roundup

- Free Shopify accounting software in 2026

- Xero alternative: the open-source path

If you want to see the $0, open-source, AI-native option run your books, book a quick demo with a co-founder or install ERPClaw and run it beside your current tool for a month. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

## Sources

- Wave, Plans and pricing (Starter and Pro plans, receipt-scanning add-on, payment processing rates)

- Zoho, Zoho Books pricing (free plan revenue threshold, user and invoice limits, autoscan allowance, paid tiers)

- Puzzle, Pricing (Starter free tier terms, AI credit allowances, paid tiers)

- Odoo, Pricing (One App Free, in-app purchase credits for digitization, per-user plans)

- GnuCash, gnucash.org (license and project scope)

Tagsfreeai-accountingtoolscomparisonopen-source
Share:XLinkedInHacker News

---

## Free Shopify Accounting Software in 2026: The Honest Guide
URL: https://www.erpclaw.ai/blog/free-shopify-accounting-software-2026/

# Free Shopify Accounting Software in 2026: The Honest Guide
A real comparison of free Shopify accounting software in 2026. What is actually free, what is freemium with a paywall, and what works for a small store.

You spent your Saturday matching Shopify payouts to your bank statement in a spreadsheet, didn’t you? The store had a good week. Forty-three orders, two refunds, a chargeback you forgot about, and one weird discount that nobody can explain. Your bank shows one number. Shopify shows a different one. For example, the gap is $87.42 and you cannot figure out where it went.

That is the Shopify accounting tax. Every store owner pays it, in time or in money. This guide is about how to stop paying it without handing $199 a month to yet another app.

We will go through every option that actually exists for free Shopify accounting software in 2026, what “free” really means in each case (spoiler: most of them are freemium with sharp paywalls), and which one fits a small store that just wants the books to balance.

## What “free Shopify accounting software” actually means

When most stores search for free Shopify accounting software, they mean one of three things:

- A tool that posts every Shopify sale, fee, refund, and payout into proper accounting books with no monthly fee.

- A free tier of a paid app that covers a small store and never asks for money later.

- A spreadsheet template that someone on YouTube swears by.

The first one is rare. The second one always has a ceiling. The third one is what you are doing now and it is why your Saturday is gone.

Below is a ranked look at the real options, what each actually costs once your store grows, and where the paywall lands.

## The five real options ranked

### 1. ERPClaw (genuinely free, open source, AI-native)

ERPClaw is what we build. It is an open-source, AI-native ERP that connects to your Shopify store, pulls in every order, fee, refund, payout, gift card, and chargeback, and posts the right entries to your books automatically. No monthly fee, no upsell tier, no premium feature gated behind a paywall. A2X, Bookkeep, Synder, and the QBO native connector are all AI-decorated (older products with chat helpers added on top); ERPClaw is the only AI-native option in this category, which means the assistant is the primary interface and the architecture was built around AI from day one. The full argument is in AI-decorated vs AI-native software.

You install it once and the Shopify integration takes over. Per-order detail goes into your books. The Shopify clearing account (the one that tracks money Shopify is holding for you) zeros out when the payout lands. Cost of goods sold posts per warehouse. Gift cards sit as a liability until they are redeemed, the way your accountant wants them to.

What you get for $0:

- Every Shopify event posted to your books, in detail

- Three-layer payout reconciliation (the math the paid apps charge $49+ a month for)

- COGS by warehouse with first-in-first-out costing

- Gift card deferred revenue done correctly

- Chargeback tracking with proper reserve entries

- The full source code on GitHub so you or your accountant can read every rule

Catch: it is self-hosted. You install it on your own machine or a small server. We have a 5-minute install walkthrough that covers it. If you can install Shopify itself, you can install this. If you want a managed version, ERPClaw Cloud Managed is our hosted offering, onboarding by request through /pricing/. The self-hosted version stays free forever under open source license.

### 2. A2X free trial (then $29 to $229 a month)

A2X is the granddaddy of Shopify accounting tools. It pioneered the clearing-account pattern that everyone now copies, including us. Their numbers are tight, their support is good, and their docs are excellent.

What it actually costs: there is no permanent free tier. You get a free trial, then you pay. Pricing tiers per a2xaccounting.com/shopify/pricing (as of June 2026) sit at roughly $29, $49, $79, $129, and $229 a month, scaling with order volume. Cost of goods sold tracking sits behind the $49 tier. Multi-store sits higher. If you grow past 1,000 orders a month, you are looking at $129+.

Use A2X if you already pay for QuickBooks Online and you need a glue layer, you have an accountant who already lives in QBO, and you do not want to think about hosting anything. Skip A2X if you want the same reconciliation math without the bill. We wrote a longer A2X comparison that goes into the line-by-line differences.

### 3. Bookkeep (cheapest paid option, $19 to $199 a month)

Bookkeep is the budget version of A2X. It posts daily summaries from Shopify into QuickBooks or Xero. Cheaper, but the granularity is lower. If your auditor or your accountant ever asks for per-transaction detail, you will be exporting CSVs.

Cost: around $19 a month (as of June 2026) to start, climbing to $199 as you add channels or want more detail. Per-transaction posting sits in the higher tier. Like A2X, it requires QuickBooks or Xero on the other end, so you are paying two bills.

Bookkeep is fine for a low-volume store whose accountant is happy with month-end summary journals. The full breakdown is on our Bookkeep comparison page.

### 4. QuickBooks Online’s native Shopify integration (technically free, often broken)

QuickBooks Online has a native Shopify connector. Inside QBO, you turn it on and it pulls Shopify data. There is no extra app fee.

The catch: QBO itself is $38 to $275 a month (as of June 2026). So “free” means free if you already pay for QBO. The bigger catch is that the native connector is famously thin. It often imports orders without the right tax breakdown, struggles with multi-currency, and tends to dump everything into a single revenue account so your gross margin reports become useless. Search the QuickBooks community forums for “Shopify reconciliation” and budget an afternoon for the rabbit hole.

If you already pay for QBO and your store does fewer than 50 orders a month, the native integration may be enough. Above that, almost every store I have talked to ends up either bolting on A2X or migrating to something else.

### 5. The hand-rolled spreadsheet (free, costs you weekends)

Almost every Shopify store starts here. You export the Shopify payout report once a week, paste it into a Google Sheet, subtract the fees, and try to make the bank deposit match. It works at five orders a week. It falls apart at fifty.

The hidden cost is your time. If you spend two hours a week on it (a typical number for stores in the 100 to 300 orders a month range), that is roughly 100 hours a year. At even a modest hourly value, the spreadsheet is the most expensive option on this list.

## The comparison table you actually want

| Option
| Real cost / month
| Per-order detail
| COGS included
| Needs another tool
| Open source
| Architecture

| ERPClaw
| $0 forever
| Yes, always
| Yes, all warehouses
| No (self-hosted)
| Yes (open source)
| AI-native

| A2X
| $29 to $229
| Daily by default
| Premium tier only
| Yes (QBO or Xero)
| No
| AI-decorated

| Bookkeep
| $19 to $199
| Higher tier only
| Limited
| Yes (QBO or Xero)
| No
| AI-decorated

| QBO native
| $38 to $275 (for QBO)
| Patchy
| Basic
| No (it is QBO)
| No
| AI-decorated

| Spreadsheet
| $0 plus your weekends
| Manual
| Manual
| No
| N/A
| None

The savings math: if you would have paid for A2X Premium at $229 a month, ERPClaw saves you $2,748 a year. If you were on Bookkeep at $199 a month, it is $2,388 a year. If you were on the spreadsheet, it is your Saturdays.

## What “free A2X alternative” actually requires

People searching for a “free A2X alternative” usually want four specific things. Here is the checklist and how ERPClaw maps to it.

- Per-transaction journal entries. Not daily summaries. ERPClaw posts every order, refund, fee, and payout as its own entry. Always.

- The clearing-account pattern. The Shopify clearing account holds money that Shopify owes you. When the payout arrives, it zeros out. If it does not zero, something is broken and you need to know. ERPClaw uses the same pattern A2X uses.

- Cost of goods sold tracking. Say you sell a $40 t-shirt that cost you $14 to source: your books need to record both numbers. ERPClaw does this per warehouse with first-in-first-out costing. A2X gates this to its $49+ tier.

- Gift card handling. A gift card sale is not revenue, it is a liability. Most cheap tools get this wrong. ERPClaw posts gift cards to deferred revenue and recognizes the revenue only when the card is redeemed.

If you want the full feature-by-feature, we keep that on the A2X comparison page. The short version: same math, no monthly fee.

## Where the “Shopify accounting open source” path wins

There is one more reason small store owners search for “Shopify accounting open source” and it is not about price. It is about not wanting a third party to disappear with your books.

A2X, Bookkeep, Synder, and the rest are SaaS. Your accounting integration lives on their server. If they raise prices, get acquired, change their API, or shut down, your books go with them. You have seen this happen with at least one Shopify app you used to love.

Open source flips that. ERPClaw runs on your machine. The SQLite file that holds every entry sits on your disk. If we vanished tomorrow, your books would still be there, every rule readable, every entry intact. That is what “you own your data” actually means.

We get into the deeper Synder vs ERPClaw differences on the Synder comparison page if you are coming from that side.

## When you should upgrade (or pay for something)

I am not going to pretend ERPClaw is right for every store on day one. Here is when you should pay for something else instead.

- You will not install anything. If running a one-line install command in Terminal is not on the table, the self-hosted option is not for you. Pay for A2X, plug it into QBO, and move on. Or wait for our managed cloud version.

- You have an accountant who only works in QuickBooks. If your accountant refuses to look at any other tool, the cheapest path is QBO plus Bookkeep. You pay for the social cost of compatibility.

- You do five orders a month. Honestly, just use a spreadsheet. The setup time on any tool is not worth it at that scale.

For everyone else (small stores doing 50 to 5,000 orders a month who would rather not pay $29 to $229 a month forever), free Shopify accounting software in the form of ERPClaw is the better answer.

## Frequently asked questions

### Is there really free Shopify accounting software in 2026?

Yes. ERPClaw is genuinely free under the GPL v3 open-source license. There is no paywall, no premium tier with the features you actually need, and no upsell. You install it on your machine and use it. The QuickBooks Online native Shopify connector is also “free” but only if you already pay $38+ a month for QBO itself.

### Is ERPClaw really free or is this a freemium pitch?

Genuinely free. The self-hosted version is open source and stays free forever. We are building a managed cloud version that will have a paid tier (because we need to pay for the servers), but the version you install yourself will never go behind a paywall. The full source code is on GitHub for anyone to read or fork.

### What is the best free Shopify accounting tool for a small store?

For a store doing under 1,000 orders a month, ERPClaw covers everything you need without a monthly bill. If you do not want to install anything, the QBO native Shopify integration works for very small stores but breaks down past 50 orders a month. Bookkeep at $19 a month is the cheapest fully-managed paid option.

### Can ERPClaw replace QuickBooks for a Shopify store?

Yes. ERPClaw is a full ERP, not a sync tool. It is the destination, not the middleware. Your books live in ERPClaw, not in QuickBooks. If you currently pay for both QuickBooks and a Shopify connector like A2X or Bookkeep, ERPClaw replaces both.

### Will I need an accountant to set up ERPClaw?

No, but we recommend running the chart of accounts past your accountant before you go live. The default chart is sensible for a small US store. If your accountant has opinions (they usually do), you can edit it. The install walkthrough covers the steps.

### What about Shopify bookkeeping free tools that are not ERPClaw?

The honest answer is that the rest of the “free” Shopify bookkeeping tools fall into three buckets: free trials of paid apps (A2X, Synder), the QBO native connector (free only if you already pay for QBO), and spreadsheet templates. None of them give you a permanent, full-featured free option. That is the gap ERPClaw fills.

## The simple ask

If you are tired of the spreadsheet weekend, here is the path:

- Read the Shopify integration page to see what ERPClaw actually does with your store data.

- Skim our pricing page to confirm there is no catch.

- Run the install walkthrough. Five minutes.

- Say “I sell through Shopify” to ERPClaw and watch it pull in your orders.

If you are still on the fence, compare us against the option you are currently paying for: A2X, Synder, or Bookkeep. The savings, in money and weekends, add up faster than you think.

You should not have to pay $229 a month to know whether your Shopify store made money this week. And in 2026, you finally do not have to.

---

## The FreshBooks Alternative for When Freelance Becomes Agency
URL: https://www.erpclaw.ai/blog/freshbooks-alternative-open-source-erp/

# The FreshBooks Alternative for When Freelance Becomes Agency
FreshBooks is great until you hire staff, take on inventory, or hit the 51-client tier. Here is the open-source FreshBooks alternative that scales with you.

You signed your 51st client this month. The FreshBooks bill jumped from $43 to $70 because you crossed a tier. You also hired your first part-time designer, started reselling a SaaS tool to two of your retainers, and your CPA mentioned at tax time that the project margin reports you have been building in Google Sheets are “creative.” You are not a freelancer anymore. You are an agency. The tool you built your billing habit around for four years is starting to feel like a cardigan that does not zip.

This is not a rant. FreshBooks is genuinely good software. If you are a solo consultant invoicing five clients a month, the workflow is clean, the iPhone app actually works, and the templates look better than anything I could draw in a weekend.

This post is for the person mid-transition. You still use FreshBooks happily for invoicing, but you are bolting things on. Toggl for the team. A second tool for the SaaS you resell. A bookkeeper to figure out staff payroll because FreshBooks Payroll only goes through Gusto. Your stack is approaching $200 a month and most of the gaps are filled by a spreadsheet you maintain at 11pm on Sundays.

ERPClaw is the FreshBooks alternative for that moment. Free, open source under GPL v3, runs on your own server, AI-native, covers invoicing plus time tracking plus inventory plus payroll plus project margin plus a real CRM. One database. No per-client tier. Keep what works in FreshBooks, replace what stopped scaling.

FreshBooks is excellent at solo invoicing, but it is AI-decorated software (a 2003 invoicing product with assistant features added later). ERPClaw is the only AI-native option in this category, which means the AI assistant is the primary interface, the spec-first architecture lets the system grow without re-pricing, and the cost base does not require per-client tiers. The full argument is in AI-decorated vs AI-native software.

## What FreshBooks does well

I want to be honest before I make the case for switching, because most “best FreshBooks alternative” listicles open with grievances and that is not a fair read.

Invoicing UX is excellent. FreshBooks treats the invoice as the center of the product. Build one in under a minute, professional templates, clients pay them without confusion. For a solo consultant, the invoice is the business.

Time tracking is built in and the mobile app is good. Start a timer on your phone between meetings, tag it to a client, have it land on next month’s invoice. The friction is low. ERPClaw has time tracking, but the UX today is a CLI command or a web form. FreshBooks wins this.

Phone photo expense capture works. Snap the lunch receipt, FreshBooks parses merchant and amount, expense lands on the right project. ERPClaw has expense capture too, but receipt scanning lives in the AI layer rather than a polished native camera flow.

Friendly to people who do not know accounting. FreshBooks never shows you a journal entry. You do not see “credit” and “debit.” You file a receipt, you send an invoice, the books balance. For a designer who became a consultant because they hated math, that floor matters.

If your business never grows past one user, five clients, and zero physical inventory, FreshBooks Lite at $23 a month is a fair price. Stay there.

## Where FreshBooks runs out

FreshBooks is built for one operator, a small client list, no physical goods. The pricing tells you so. From freshbooks.com/pricing (as of June 2026):

- Lite: $23/month, 5 billable clients

- Plus: $43/month, 50 billable clients

- Premium: $70/month, unlimited clients

- Select: custom pricing, revenue over $250K

The client tier is the whole story. Cross 50 clients and the bill almost doubles. Hire staff and you bolt on Gusto, because FreshBooks Payroll is a Gusto integration. Resell anything physical (a swag box, hardware, a SaaS license) and you discover FreshBooks has no inventory of any kind.

The list of things FreshBooks does not do, that growing service businesses end up needing:

- No real inventory. If your agency resells SaaS subscriptions or kits, you cannot track stock or cost of goods sold. People work around this in spreadsheets.

- No native payroll for staff. Gusto integration only. Once you have three employees, you are paying for two systems.

- CRM is shallow. Client contacts, yes. A sales pipeline with lead scoring, proposals attached to opportunities, deal stages? No. Agencies hitting ten people end up paying HubSpot Starter at $50 a month.

- Project P&L is per-client billing, not true margin. You see what you billed a client. You cannot easily see staff hours times loaded labor cost minus pass-through expenses equal to actual gross margin. That gap is where bad client decisions get made.

- No multi-entity. Second LLC? Second subscription.

- Multi-currency is Premium tier. Bill a UK client in pounds, you are on the $70/month plan minimum.

None of these are bugs. They are deliberate scope choices. FreshBooks just chose a smaller surface than your business is starting to need.

## The real monthly bill

The headline FreshBooks number is $23 to $70. The actual bill for a five-person agency that started on FreshBooks looks like this:

- FreshBooks Premium, $70

- Gusto payroll, 5 employees, $40 base plus $6 per person = $70

- Toggl Track, 5 seats at $9 = $45

- HubSpot Starter sales pipeline, $50

- Spreadsheet plus fractional bookkeeper for the gaps, $300 to $600

That is $535 to $835 a month before any of it talks to any of the others. None of these tools share a database. Toggl entries land in FreshBooks via CSV import. The HubSpot deal that became a client does not auto-populate the FreshBooks customer. The Gusto payroll runs do not land on a project margin report because there is no project margin report.

Starting price is fine. End-state stack is not. If your monthly tool bill across FreshBooks plus add-ons has crept past $300, this post is for you.

## What ERPClaw replaces

A typical FreshBooks alternative comparison stays in the same category: Wave, Zoho Invoice, Bonsai, Harvest. Same shape, slightly different price.

ERPClaw is a different category. Invoicing is one domain among many in a single system that shares one database. You stop running a stack and start running a system:

- Invoicing: estimates, invoices, recurring, late fees, Stripe online payments. Same shape as FreshBooks; behind it is a real audit-grade ledger your CPA will recognize.

- Time tracking: timesheets per employee, per project, per task. Less polished than the FreshBooks mobile timer; more useful at agency scale because it ties to project margin and payroll in the same database.

- Expense capture: receipt upload, AI parse, attach to project.

- Inventory: items, warehouses, batches, serial numbers, reorder levels. If you resell anything, this exists.

- Payroll: native US payroll with FICA, federal and state withholding, W-2 generation, garnishments. Not a Gusto integration. Same database as your time entries and invoices.

- CRM: leads, opportunities, pipelines, contacts. Enough to stop paying for HubSpot Starter at $50 a month.

- Project P&L: every time entry has a labor cost. Every expense has a project tag. Project margin is a query, not a spreadsheet. See features.

- Multi-entity in the core: one install, N companies, intercompany and consolidation included. No second subscription.

- Stripe deep integration: all core Stripe actions including ASC 606 revenue recognition for any agency selling a productized service or a retainer with deferred revenue. Detailed in our Stripe launch post.

Total those as separate SaaS subscriptions and you are past $1,500 a month for a 10-person agency. ERPClaw is $0 forever (open source), self-hosted on your own infrastructure.

This is not just a price argument. It is structural. When invoicing, time, payroll, and CRM share one database, your project margin report knows actual labor cost and your AR aging report knows the deal stage that produced the customer. No sync to break.

## Side by side

The full comparison lives at /compare/freshbooks. Pulling the table here:

| Feature
| FreshBooks
| ERPClaw

| Price
| $23 to $70/month
| Free (open source license)

| Best for
| Freelancers, small services
| Solo to enterprise

| Invoicing
| Excellent
| Excellent (full ledger behind it)

| Time tracking
| Built in (great mobile UX)
| Built in (less polished UX)

| Inventory
| Not available
| Built in

| Multi-currency
| Premium tier
| USD in v1

| Per-client pricing
| Yes (5/50/500 tiers)
| Unlimited, always

| Self-hosted
| No
| Yes

| Open source
| Proprietary
| open source

| AI capabilities
| AI-decorated (chat features added on top)
| AI-native (assistant is the primary interface)

| Database
| FreshBooks cloud only
| SQLite or PostgreSQL via PyPika, on your machine

Two rows deserve a second look.

Per-client pricing. This is the FreshBooks moment that triggers the FreshBooks alternative search more than any other. You signed your 51st client and the bill nearly doubled. Cross some other invisible threshold and Select pricing is “let us hop on a call.” ERPClaw does not gate by clients, users, employees, entities, or transactions. You install it once.

AI-native. FreshBooks bolted on a few AI features. ERPClaw is designed to be operated through conversation: Telegram, WhatsApp, the web dashboard, or any OpenClaw client. For example: “Send Acme a $5,000 invoice for April retainer, net 30, and remind me if it is unpaid by May 15.” It writes the entry, posts it, and confirms. Not a chatbot wrapper around forms.

## Where FreshBooks still wins

I promised honest. Places where FreshBooks is the better tool today:

Mobile-first solo workflow. If your job is “phone in one hand, coffee in the other, send an invoice between meetings,” the FreshBooks iPhone app is better than anything ERPClaw ships in 2026. We have a web dashboard and chat interfaces. We do not yet ship a native mobile app at FreshBooks polish.

Out-of-box invoice template aesthetics. FreshBooks has put years into invoice design. The default ERPClaw invoice PDF is clean, but a FreshBooks template the day you install looks more polished.

Expense receipt scanning in two taps. Photograph the lunch receipt, done. ERPClaw expense capture works through file upload plus AI parsing. Same result, higher friction.

You truly never want to think about a server. ERPClaw is self-hosted by default. If “no servers, ever” is a hard line, FreshBooks handles that, or have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

If all four describe you and your business is staying solo, stay on FreshBooks. Bookmark this post for the year you hire your first employee.

## When to make the switch

The signals that you have outgrown FreshBooks and the FreshBooks alternative search is justified:

- You crossed the 50-client tier and the bill jumped to $70, and you are buying client slots you do not need just for headroom.

- You hired your first part-time staff and are now running Gusto separately.

- You started reselling something physical (kits, swag, hardware) or a SaaS subscription you bill clients for.

- You use a separate tool (HubSpot, Pipedrive, Notion) to track sales because FreshBooks contacts are not a CRM.

- Your CPA asked for a project-level P&L and you sent her a spreadsheet you built by hand.

- You spun up a second LLC and FreshBooks wants you to buy a second subscription.

Three or more = switch. One or two = bookmark this and revisit in six months.

## How a migration actually works

A standalone “Migrating from FreshBooks to ERPClaw” guide is in our docs roadmap for Q3 2026. Short version:

- Export from FreshBooks. Settings, Account Settings, Export. Zip with clients, invoices, expenses, time entries, items. About 5 minutes.

- Pick a cutover date. Start of a month. Close FreshBooks at the end of the prior month, open ERPClaw with opening balances on day 1.

- Import master data. Clients, items, recurring invoice templates. One-line imports, about 15 minutes.

- Import opening balances. FreshBooks AR aging loads as opening invoices. Bank balance loads as opening cash. The validation engine rejects any imbalance, so you know the import is clean.

- Send your first invoice from ERPClaw. Day 1 of the new month, to a client that has paid reliably for years. Feel the workflow on a low-stakes invoice first.

- Run parallel for one cycle. Recommended for agencies. Send next month’s invoices from both systems for one week, compare, then kill FreshBooks.

Solo consultant under 50 clients: a Saturday afternoon. 10-person agency with active retainers: two weekends.

## FAQ

Is there a free FreshBooks alternative?

Yes. ERPClaw is free under GPL v3 and self-hosts on your own infrastructure. The lowest FreshBooks plan is $23/month for 5 clients. “FreshBooks alternative free” is either ERPClaw self-hosted or Wave at $0 with a much smaller feature set.

Is ERPClaw really good for agencies, not just freelancers?

Yes, and that is the wedge. FreshBooks shines for solo. ERPClaw works for solo and keeps working when you hire, take on inventory, run a sales pipeline, or open a second entity. Agency build at /for/agency; solo build at /for/solo-founder.

FreshBooks vs ERPClaw for someone who only invoices five clients?

Stay on FreshBooks Lite. $23 a month is a fair price for the polish, the mobile app, and the receipt scan. Switch when you cross the tier.

Best invoicing software for freelancers in 2026?

Honest answer: FreshBooks if you are staying solo. ERPClaw if you are trending toward an agency, hiring, or selling anything beyond services.

Will my CPA accept reports from ERPClaw?

Yes if your CPA understands a trial balance and a ledger export, which is the same skill they use for non-FreshBooks clients. ERPClaw exports a clean trial balance, P&L, balance sheet, and detailed ledger in CSV or PDF. If your CPA only knows FreshBooks, ask before you switch.

What happens if ERPClaw disappears?

You still have the software. open source license, your code, your data, your server. The repo at github.com/avansaber/erpclaw is forkable. Compare to a SaaS sunset where the vendor pulls the plug and you have a CSV export and 30 days.

## Try it

Install takes five minutes. The whole ERP on your machine:

`clawhub install erpclaw`
From there, talk to it: “I run a 6-person design agency, set me up.” It picks the right modules, generates the chart of accounts, and is ready for invoicing.

If you want to read first:

- /features for the module breakdown

- /pricing (it is free)

- /compare/freshbooks for the side by side

- /for/agency for the agency pitch

- /for/solo-founder if you are still a one-person shop

- /docs for install and the migration guide

Repo: github.com/avansaber/erpclaw. Star it to follow along; we ship weekly. Email `support@avansaber.com` if you hit something broken.

FreshBooks is the right software for solo consultants who plan to stay solo. ERPClaw is the right software the day that stops being true. If you crossed the 50-client tier this year, hired anyone, or sold anything you had to ship, you are already on the other side of that line.

Time to install.

---

## From SAP Consultant to Open Source ERP Builder
URL: https://www.erpclaw.ai/blog/from-sap-consultant-to-open-source-builder/

# From SAP Consultant to Open Source ERP Builder
18 years across enterprise software, six of them rolling out SAP at Accenture. Here is why I left, what I learned, and why ERPClaw exists now.

In 2009, in a fluorescent-lit conference room in Greensburg, Pennsylvania, I watched a $200-an-hour SAP consultant explain the Vendor Master configuration screen to an Allegheny Power accountant who had been doing accounts payable for thirty-one years.

She already knew the answer. She had known it for a decade. She was waiting, politely, for the consultant to finish so she could nod and go back to her desk and approve the invoice that had been sitting in her queue for two days because the workflow we had configured was wrong.

I was the consultant’s junior. My billable rate was lower. I was twenty-four years old and I had been at Accenture for less than a year. I remember thinking, very clearly, two things at once. First, that this was the most expensive way to discover something that the person sitting across the table already knew. Second, that I was going to do this for at least the next five years because the project plan said so.

That moment is the reason ERPClaw exists. It just took seventeen years to get from there to here.

## The Accenture years

I spent six years at Accenture rolling out SAP for utility companies. Allegheny Power. E.ON in the UK and Germany. American Water. The work was good. The clients were serious people running real infrastructure that millions of households depend on. I learned more about the actual mechanics of how a large business runs in those six years than I have learned in any other six-year window of my life.

I also learned what enterprise software costs.

The published license fee for an SAP S/4HANA seat is one number. The real cost is something between three and ten times that number, depending on who you ask. The implementation partner bills hours. The change management team bills hours. The training team bills hours. The integration team bills hours to talk to the data team that bills hours to talk to the security team. A mid-sized utility writes checks for tens of millions of dollars before a single accountant can post a journal entry.

I am not going to pretend SAP is bad software. It is not. It runs the spine of global commerce and it has earned that position. What I will say is this: the gap between what SAP actually does for a business and what the business pays to get it has been growing for thirty years, and nobody on the buying side has had a real alternative.

The thing the consultant was explaining to the accountant in Greensburg was not complicated. It was a screen with maybe twelve fields. The reason it was expensive was not the screen. It was the cost of getting the screen, configured correctly, in front of her, in a system that her company had committed to for the next fifteen years.

I started writing code on the side around year four. Mostly utilities. A small reporting tool. A spreadsheet macro that saved a client team six hours a week. Nothing serious. But the contrast was sharp. Six hours of consulting time was around $1,200. The macro was an afternoon and it ran forever for free.

I started thinking about what it would take to build software that did the work of an SAP module without the implementation overhead. I did not have an answer in 2012. I did not have one in 2014, either. I left Accenture in 2015 and started building.

## The indie SaaS years

Between Accenture and ERPClaw there is a decade I do not talk about much. It is not a secret. It just gets compressed in my head into a single phrase: “I was building inventory software.”

I built ZapInventory because I kept getting asked by small wholesalers and Shopify merchants to help them figure out their stock. They were on QuickBooks. QuickBooks does not really do inventory. They were on spreadsheets. Spreadsheets do not really do inventory either. They needed something in between QuickBooks and a $50,000 NetSuite implementation, and almost nothing existed in that gap.

ZapInventory grew. By 2020, Crozdesk ranked it the third inventory management software in the world, ahead of SAP Business One and Zoho Inventory. That sentence still feels slightly absurd to type. I was a solo founder running a small team out of Pune. We were ahead of SAP on a list that mattered to buyers. Not because we were better than SAP. We were not. We were ahead because for the buyer who came to that list, we were the right thing and SAP was the wrong thing, and the list was honest about that.

ZapInventory was acquired by InvenSync Inc. I built a few other things. StockVR, a 3D warehouse visualization tool, got mentioned by Allied Market Research in 2024 in the same paragraph as Oracle, IBM, and SAP. SuperSnail, a mobile game my team made on the side, crossed a million downloads and got acquired by QCPlay Digital. I became an IEEE Senior Member somewhere along the way, published a couple of papers in Springer proceedings on AR and inventory visualization. The CV grew. The lessons compounded.

The lesson that stuck, the one that mattered for what came next: the default ERP for a small business should not cost $50,000 to install. It should cost approximately zero, and it should know what it is doing.

In 2024 I did not yet think that was possible. By the end of that year I started to think it was.

## The thing that changed

Two things changed.

The first was AI. Not the chatbot kind. The “I can describe what I want and the software writes itself” kind. By late 2024 I was writing more code per week with AI than I had written by hand in the previous three years combined. The cost of building software, on the maker side, fell off a cliff. The cost of buying software, on the buyer side, did not.

That gap is an arbitrage. It is the kind of arbitrage that does not last. Either buyers stop paying or sellers start cutting prices. Both are happening, slowly, in pockets. ERPClaw is a bet that the cleanest way through the arbitrage is to build the new default from scratch, AI-native, free, open source, and let the gap collapse on its own.

The second thing that changed was OpenClaw. OpenClaw is an AI-native skill execution layer. You install a skill, you talk to your assistant in natural language, the assistant calls the right action under the hood. It is not a chatbot bolted onto a CRUD app. It is a runtime where the skill is the unit of distribution and the AI is how the user interacts with it. Once OpenClaw existed, the architecture for what I had been wanting to build for a decade clicked into place in a weekend.

ERPClaw is the answer to the question I have been carrying since that conference room in Greensburg. What if the accountant just talked to the system? What if the system did not require a $200-an-hour consultant to explain a screen? What if the screen, in fact, did not really need to exist?

## What ERPClaw actually is

In concrete terms, ERPClaw is a complete double-entry ERP that installs in one command, runs locally on a SQLite database, and is driven by natural language through your AI assistant.

It covers the full back office. Core accounting, sales, purchasing, inventory, HR, payroll, manufacturing, projects, assets, CRM. Vertical packs for healthcare, education, property management, retail, construction, food service, hospitality, legal, nonprofit, and more. Deep integrations with Stripe and Shopify that post per-transaction journal entries directly into your general ledger. ASC 606 revenue recognition. ASC 842 lease accounting. W-2 generation. Multi-entity consolidation.

It is open source. It is free. It runs on your laptop or your own server. Your data lives on disk in a single file you own. The full source is on GitHub. If you want to fork it, fork it. If you want to install it, the docs are public.

There is also a self-extending layer called ERPClaw OS. It is the part of the system that lets AI agents generate new vertical modules within constitutional financial laws. We have used it to ship three modules end-to-end with no human-written code: groomingclaw, tattooclaw, storageclaw. They pass the same full constitution-enforced test suite the hand-written modules pass. The constitution does not let them post bad GL entries. That is the part of the architecture I am proudest of, and it is the part I think will matter most over the next five years.

You can see the full feature set at /features, and the pricing is on /pricing. The pricing page is short. The product is free. The hosted version, when it ships later this year, will be priced at the cost of running it for you, with a generous free tier.

## The bet

The bet I am making with ERPClaw is that AI-native is not a feature. It is the new default architecture for business software, and the next ten years are going to look like the migration from on-premise to SaaS that happened between 2008 and 2018. Same shape, faster timeline.

A SaaS ERP that bolts a chatbot onto a 1995 data model is not the answer. The answer is a system designed, from the database layer up, to be driven by an AI agent on behalf of a human who does not want to learn the screen. The screen, where it exists, is a fallback. The agent, where it works, is the primary interface. The data model has to be clean enough that the agent does not lie. The audit trail has to be tight enough that the CFO can sign the books. The license has to be open enough that the buyer never has to worry about lock-in.

Those four constraints rule out almost every existing ERP. They also rule out almost every “AI accounting” startup I have looked at, because most of them are AI bolted onto QuickBooks, which means QuickBooks is still the bottleneck.

I do not know if ERPClaw wins. I know the category is going to exist. I have spent eighteen years across enterprise software, including six years inside the consulting machine that built the previous category, and I have a strong opinion about what the new one looks like. ERPClaw is that opinion, expressed as code.

## Where to go from here

If you read this far and you want to actually try it, the easiest path is one command:

`clawhub install erpclaw`
That installs the core ERP. Then you talk to your assistant. Tell it what your business is. It will set up your chart of accounts, your fiscal year, your tax rates, and suggest the modules you actually need. The whole thing takes about three minutes.

If you want to read first, the docs are a good place to start. If you want to follow along, the repo is at github.com/avansaber/erpclaw. A star helps. So does an issue. So does a pull request.

If you want to know more about the company behind it, that is AvanSaber. If you want to know more about me personally, my LinkedIn is the easiest way to get there, and I read every message that does not start with “circling back.”

The accountant in Greensburg retired a few years ago. I never got to tell her that her patience that day, sitting through the explanation she did not need, became the seed of a thing she would have actually used. I think she would have liked it. I think she would have ignored most of it and gone straight to the AP screen and posted the invoice in eight seconds without asking anyone for permission.

That is the ERP I want to build. That is the ERP I am building.

---

## Shopify Gift Card Accounting: the Deferred Revenue Trap
URL: https://www.erpclaw.ai/blog/gift-card-deferred-revenue-shopify/

# Shopify Gift Card Accounting: the Deferred Revenue Trap
Gift card sales aren't revenue, they're a liability. Here's how Shopify gift card accounting actually works, why most stores get it wrong, and the fix.

Suppose you sold $5,000 in gift cards last December. Big holiday push, lots of “she’ll pick what she wants” purchases. You closed the year feeling great. Then in late January your CPA calls and asks why your books look weirdly profitable, and why the same thing falls apart the second week of February when redemptions start landing. Your margins go from “best year ever” to “wait, did I actually make money?” in about ten days.

That whiplash isn’t random. It’s what happens when gift card sales get booked as revenue on the day the card is sold instead of on the day the card is used. This is one of the most quietly mishandled items in retail accounting, and it bites small Shopify stores harder than almost anything else because nobody flags it for you. The Shopify dashboard shows the gift card sale as part of your revenue. Your bank account agrees. Everything looks normal until tax season, when it doesn’t.

This post covers what gift card sales actually are in accounting terms, the dollar math, what stores get wrong, ASC 606 in plain English, how state laws complicate expiration, and how to stop tracking this by hand.

## What a gift card sale actually is (it’s not revenue)

Say a customer buys a $100 t-shirt: you took their money and gave them a t-shirt. Sale done. You earned the revenue. Your only remaining job is to not have the shirt fall apart.

For example, when a customer buys a $100 gift card, something different happened. You took their money. You gave them a piece of plastic (or an email) that promises you’ll hand over $100 worth of stuff later. You haven’t actually delivered anything yet. The customer can walk in tomorrow, six months from now, or never. Until they redeem it, you owe them $100 worth of inventory, service, or both.

In accounting language, that “I owe you something” relationship is called a liability. A gift card sale moves cash into your account, but it doesn’t create revenue. It creates a promise to deliver. Revenue only shows up when you fulfill that promise.

The technical name for this kind of liability is deferred revenue, sometimes called “unearned revenue” or “customer deposits.” All three names mean the same thing: money held until used. You have the cash, but you haven’t earned it yet. It belongs to the customer in the form of a future obligation.

This is the same accounting bucket that holds annual subscription prepayments, customer deposits on custom orders, pre-orders before shipping, and law firm retainers. Gift cards are the retail version. Cash in, service not delivered, liability on the books.

## The journal entry math, with real dollars

For example: let’s walk through the actual numbers on a $100 gift card sale and a later redemption. No accounting jargon, just what each line means.

### Sale day: gift card sold

Say a customer buys a $100 gift card on December 15. Cash hits your Shopify account.

In your books:

- Cash (or Shopify Clearing) goes up by $100. You have the money.

- Gift Card Liability goes up by $100. You owe a $100 promise.

Notice what didn’t change: revenue. Your sales total for December didn’t move when this gift card was sold. The cash showed up, the liability showed up, and they cancel each other out on the balance sheet. You’re not richer. You’re just holding someone else’s money.

If you want this in formal debit and credit format:

`DR Cash / Shopify Clearing $100
CR Gift Card Liability $100`

### Redemption day: customer redeems the card

Suppose the same customer comes back in February, picks out an $80 sweater, and pays with the gift card. They walk out with the sweater. Their gift card balance is now $20.

Two things just happened. The customer received the goods, so revenue is finally earned. And your obligation to that customer dropped by $80.

In your books:

- Gift Card Liability goes down by $80. You owe $20 less than you did this morning.

- Revenue goes up by $80. You finally earned it.

- Inventory goes down by the cost of the sweater (let’s say $32).

- Cost of Goods Sold goes up by $32.

`DR Gift Card Liability $80
CR Revenue $80
DR Cost of Goods Sold $32
CR Inventory $32`
The $20 left on the card is still a liability. It sits on your balance sheet until the customer redeems it, lets it expire, or it gets escheated to the state (more on that in a minute).

After both transactions, your December P&L shows zero revenue from this customer. Your February P&L shows $80. That matches what actually happened. The customer didn’t buy anything in December. They prepaid for a future purchase.

## What most Shopify stores get wrong

Here’s what almost every small store does instead. It’s an honest mistake and Shopify doesn’t help you avoid it.

### Mistake 1: Booking the full gift card sale as revenue on day one

The store sees the $100 hit their bank account, the Shopify dashboard shows a sale, and they book it as revenue right there. December looks like a great month. The accountant has nothing to flag because the bank deposit and the books agree.

The trouble is that February also gets booked as revenue when the gift card is redeemed, because the redemption shows up as a sale in the dashboard too. Now the same $80 was counted twice: once when the gift card was sold, again when it was redeemed. Revenue is overstated by exactly the redemption amount.

### Mistake 2: Under-reporting the gift card liability

Because revenue was already booked in December, no liability ever showed up on the balance sheet. The store carries no record that they owe customers anything. This breaks the balance sheet. It makes the business look more solvent than it actually is. And if you ever try to sell the business, raise money, or get a bank loan, the buyer’s accountant will spot it in five minutes and the deal slows down.

### Mistake 3: Over-reporting Q4 profit

December is peak US gift card volume. If you booked all of it as revenue, your Q4 P&L is inflated by the full unredeemed balance. You pay tax on profit you didn’t earn, set next year’s targets on a false baseline, and have a brutal Q1 when the same dollars get “earned” again on redemption while you’re fulfilling goods at a real cost. The pattern is consistent: a great Q4, a confusing Q1, and an angry CPA.

### Mistake 4: Forgetting about breakage

Some gift cards never get redeemed. People lose them or forget about them. The unredeemed portion is called breakage, and what you do with it depends on your state (covered below). The trap: if the original sale was already booked as revenue, the liability never existed in your books, so there’s nothing to track or release.

## What ASC 606 says, in plain English

ASC 606 is the US accounting standard that governs revenue recognition. It applies to every business that sells goods or services, not just public companies. The full text is dense, but the part that matters for gift cards fits in one sentence.

Revenue is recognized when the performance obligation is fulfilled.

A “performance obligation” is the promise you made when the customer paid. For a t-shirt sale, the obligation is fulfilled when you hand over the shirt. For a gift card sale, the obligation is fulfilled when the customer redeems the card and walks out with goods.

So under ASC 606, gift card revenue is recognized at redemption, not at sale. The day the card is sold, you’re holding cash for an unfulfilled obligation. The day the card is redeemed, you’ve fulfilled the obligation, and only then can you call it revenue.

This is why the journal entry pattern above (cash up + liability up on sale, liability down + revenue up on redemption) isn’t optional or aggressive. It’s the standard. Your accountant will follow it. Your auditor will require it. Your tax return assumes it. The only question is whether your bookkeeping system follows it automatically or whether you’re going to fix it by hand every quarter.

For the deeper version, the AICPA’s revenue recognition resource page walks through the five-step ASC 606 model. For gift cards, the one-sentence version is enough.

## The expiration and escheatment problem (state laws matter)

What happens to gift cards that never get redeemed? Two questions: can you keep the money, and when can you call it revenue? The answer depends on your state.

### Federal rule: cards can’t expire for at least 5 years

The CARD Act of 2009 says retail gift cards can’t expire for at least five years from issuance, and can’t carry monthly fees for the first year. You can’t print “expires in 12 months” on a gift card and refuse to honor it.

### State rule: escheatment laws can send the money to the state

Once a gift card is “abandoned” by your state’s definition (usually 3 to 5 years of inactivity), most states require you to remit the unredeemed balance to the state’s unclaimed property fund. This is called escheatment. The state holds the money in case the customer shows up. You don’t keep it.

Delaware, New Jersey, New York, and Massachusetts actively audit retailers for unremitted gift card balances. Others (Florida, Arizona, a few more) exempt gift cards from escheatment entirely, which means breakage stays with you and can eventually be recognized as revenue using a historical redemption rate (most retailers use 5 to 15 percent based on actual data). The National Conference of State Legislatures maintains a gift card law summary by state worth bookmarking.

The point isn’t that you need to memorize 50 sets of rules. The point is that the gift card liability on your balance sheet isn’t pure profit waiting to happen. Some of it belongs to the state. Until you know which is which, leaving it on the liability line is the safe call.

## Tools that handle this for you (and which ones don’t)

Here’s the honest landscape for Shopify gift card accounting tools as of 2026.

Shopify itself doesn’t separate the books for you. Shopify’s reporting shows gift card sales mixed into your revenue dashboard, and redemptions show up as part of the order total. The gift card balances sit in a separate area of the admin, but they don’t sync to your accounting system in a way that produces correct journal entries. You’re on your own to fix the bookkeeping.

A2X handles gift cards on its premium tier ($79-229/month). The basic tier lumps gift card revenue into general sales, which means most A2X users on lower tiers are still booking gift card sales as immediate revenue without realizing it.

Bookkeep, Synder, Webgility all support gift card deferred revenue with various caveats. Bookkeep handles it cleanly on its standard plan. Synder requires custom rules. Webgility supports it but the setup is fiddly.

ERPClaw is the AI-native option in this list. Gift card deferred revenue is included on the free tier with no setup: sale posts to the Gift Card Liability account, redemption reverses the liability and posts revenue, and breakage tracking sits on top of the same engine. The full GL rule set for the Shopify integration is open-source, and the Shopify architecture doc shows every journal entry. The A2X comparison walks through the feature-by-feature differences and the pricing page covers the rest.

## FAQ

### Is selling a gift card considered revenue under GAAP?

No. Under both GAAP and ASC 606, a gift card sale creates a liability called deferred revenue (or “gift card liability”) on the balance sheet. Revenue is recognized only when the customer redeems the card and you deliver goods or services. The cash from the sale goes on the books, but it offsets the liability and doesn’t show up on the income statement.

### How do I record a gift card sale on Shopify?

Two journal entries. On the sale, debit Cash (or Shopify Clearing) and credit Gift Card Liability for the full sale amount. On redemption, debit Gift Card Liability and credit Revenue for the redeemed amount, plus the usual COGS entry for the inventory delivered. ERPClaw posts both entries automatically when its Shopify integration is connected. The Shopify payout reconciliation guide covers how the related payout accounting flows.

### What is the difference between a gift card sale and a gift card redemption in accounting?

Sale = cash in, liability up, no revenue. Redemption = liability down, revenue up, COGS recorded. Two different transactions, two different sets of journal entries. Mixing them up is the single most common mistake in retail bookkeeping.

### What happens to gift cards that are never redeemed?

It depends on your state. In states with strong escheatment laws (Delaware, New Jersey, New York, and others), unredeemed gift card balances must eventually be remitted to the state’s unclaimed property fund after a dormancy period. In states that exempt gift cards from escheatment (Florida, Arizona, a few more), you can recognize the breakage as revenue using a historical redemption rate, typically 5 to 15 percent. Federal law (the CARD Act) prevents gift cards from expiring for at least five years regardless of state.

### Does Shopify automatically handle gift card deferred revenue?

No. Shopify reports gift card sales as dashboard revenue and shows redemptions as normal orders. It doesn’t produce the journal entries to defer revenue properly. You need either a middleware app (A2X premium, Bookkeep, Synder) or an integrated tool that handles deferred revenue.

### When can I recognize gift card breakage as revenue?

Only if your state allows it. In states that exempt gift cards from escheatment, you can recognize breakage proportionally as redemptions occur, using a historical breakage rate based on at least two years of data. In escheatment states, the unredeemed balance gets remitted to the state instead. Talk to a CPA familiar with your state’s unclaimed property rules before you book any breakage.

## Stop overstating your December

If your store sells gift cards, the difference between “great Q4” and “honest Q4” is one journal entry pattern, applied consistently. The accounting standard has been clear since 2018. The state escheatment rules have been around for decades. The only thing that’s been missing is bookkeeping software that handles all of this without charging you $79 to $229 per month for the privilege.

Whatever tool you pick, the rule is the same: gift card sales create a liability, not revenue. Revenue shows up at redemption. Breakage waits for state rules. Get those three right and your December stops lying.

Install the ERPClaw Shopify integration for the free version of this, read the Shopify architecture docs, or check the A2X comparison page and the pricing page to see what’s included. The companion piece on Shopify payout reconciliation covers the other half of the picture.

---

## FIFO vs Weighted Average Inventory: COGS and Profit
URL: https://www.erpclaw.ai/blog/inventory-fifo-vs-weighted-average/

# FIFO vs Weighted Average Inventory: COGS and Profit
Compare FIFO and weighted average inventory costing with one rising cost example, a decision guide, and the effects on COGS, profit, tax timing, and ending inventory.

You bought 100 units of a product at $10 each in January and another 100 at $15 each in March. By April you sold 150 units. FIFO reports $1,750 in cost of goods sold and $750 in ending inventory. Weighted average reports $1,875 in cost of goods sold and $625 in ending inventory.

In this rising cost example, FIFO reports lower COGS and higher gross profit for the period. Weighted average blends the two purchase costs and reports a smoother cost per unit. Both methods eventually recognize the same $2,500 total purchase cost if every unit is sold and there are no write-downs or other cost adjustments.

Use this quick comparison before the full walkthrough:

- FIFO: $1,750 COGS, $750 ending inventory, higher current gross profit in this example, and a typical fit for perishable or lot-tracked stock.

- Weighted average: $1,875 COGS, $625 ending inventory, smoother unit cost, and a typical fit for high-volume identical SKUs.

## What inventory costing actually solves

Every product you sell has a cost. The cost shows up on your P&L as Cost of Goods Sold (COGS), which is what you subtract from revenue to get gross margin. The cost is also what shows up on your balance sheet as the value of inventory you still hold.

The trouble is that you bought the same product at different prices over time. Maybe your supplier raised prices in March. Maybe you got a volume discount on a big buy. Maybe shipping was higher on the second order. So when you sell a unit in April, the question is: which unit did you sell? The cheap one from January or the expensive one from March?

The answer determines how much COGS lands on your P&L this period and how much inventory value remains on your balance sheet. Different methods give different answers. Inventory costing is the rules for picking which cost to assign.

There are four methods: FIFO (first in, first out), LIFO (last in, first out), weighted average, and specific identification. LIFO is allowed in the US but banned under IFRS, so most international businesses cannot use it. Specific identification is for high-value unique items (cars, jewelry). For most small businesses, the choice is FIFO or weighted average.

## FIFO, walked through with real numbers

FIFO assumes you sell the oldest inventory first. The unit you bought in January gets shipped before the unit you bought in March. The COGS reflects the cost of the oldest units you have on hand.

Example. You bought 100 units at $10 in January, 100 units at $15 in March, and you sold 150 in April.

Under FIFO, you sold:

- 100 units from January at $10 = $1,000

- 50 units from March at $15 = $750

- Total COGS: $1,750

- Inventory remaining: 50 units at $15 = $750

The number on your P&L for this period is $1,750 in COGS. The number on your balance sheet for ending inventory is $750.

The mental model: FIFO is the “first-in line is first-out the door.” Whatever you bought first is what you assume you sold first.

## Weighted average, walked through with real numbers

Weighted average computes one blended cost per unit across all your purchases, then applies that blended cost to every unit sold. Same example.

Same example. 100 units at $10 in January, 100 units at $15 in March, sold 150 in April.

Total cost of all units: 100 × $10 + 100 × $15 = $1,000 + $1,500 = $2,500
Total units: 200
Weighted average cost per unit: $2,500 / 200 = $12.50

Sold 150 units at the weighted average:

- COGS: 150 × $12.50 = $1,875

- Inventory remaining: 50 × $12.50 = $625

The number on your P&L for this period is $1,875 in COGS. The number on your balance sheet for ending inventory is $625.

Notice how the two methods produce different numbers. FIFO COGS was $1,750. Weighted average COGS is $1,875. The difference is $125 in this single example. Across thousands of transactions, the total numbers diverge meaningfully.

## The same total cost, different timing

Here is the part that surprises most people. If the inventory is fully sold and there are no write-downs or other cost adjustments, FIFO and weighted average produce the same total cost. The difference is timing. FIFO recognizes the cost in a different period than weighted average does.

In our example, the full $2,500 of inventory cost flows through COGS eventually. FIFO put $1,750 of it in this period and $750 in the next. Weighted average put $1,875 in this period and $625 in the next. After the inventory is fully sold, both methods total $2,500 in COGS. Same number.

This matters because COGS in any given period directly affects:

- Your gross margin in that period

- Your taxable income in that period

- Your tax bill for that year

Different timing means different tax in any given year. Across many years it evens out.

## Which one shows higher profit when prices are rising?

Inventory costs almost always rise over time (inflation). When prices rise:

- FIFO assumes you sold the cheap stuff first. Your COGS uses the older lower costs. Your COGS is lower. Your gross margin is higher. Your taxable income is higher. Your tax bill is bigger.

- Weighted average blends old and new. Your COGS sits in the middle. Your gross margin is in the middle. Your tax bill is in the middle.

So during an inflationary period, FIFO makes your business look more profitable but you pay more tax. Weighted average looks less profitable but pays less tax.

A lot of business owners pick FIFO because their accountant said so, then complain about the tax bill, then realize their accountant picked it specifically because the higher reported profit looked better to a bank for a loan application. Pick on purpose.

## Tax implications in detail

The IRS allows either method. Once you pick one, you stick with it for that product line; switching mid-year requires IRS Form 3115 and a multi-year adjustment.

FIFO tax consequence. In a rising-price environment, FIFO produces higher taxable income because COGS is lower. You pay more tax this year. The “extra” tax is essentially a prepayment, since over the full lifecycle the total tax is the same.

Weighted average tax consequence. Smoother taxable income year to year. Lower tax in the period your costs are rising, higher tax later if costs eventually fall. Less volatile tax bills.

For most US small businesses on accrual basis, the weighted average method produces a simpler and slightly more tax-favorable result during inflation. This is one reason many ERPs default to weighted average.

## Inventory valuation on the balance sheet

The other side of the COGS coin is ending inventory value, which sits on your balance sheet as a current asset.

- FIFO values ending inventory at the most recent prices. Inflation pushes the balance sheet inventory value higher. Looks better in a balance sheet ratio analysis.

- Weighted average values ending inventory at the blended cost. Lower than FIFO during inflation, higher during deflation.

A lender looking at your balance sheet sees a different inventory number depending on which method you use. For most small businesses this does not matter day to day, but if you are applying for an asset-backed loan, the choice meaningfully affects how much you can borrow against inventory.

## When FIFO is the right pick

Perishable goods. If you sell food, beverages, dairy, fresh products, FIFO matches reality. The oldest inventory really does need to go out first or it spoils. Your books match what is happening in the warehouse.

High-value low-turn items. If you sell expensive items that move slowly, FIFO gives you a clean lot-by-lot trail of which specific batch a sale came from.

You want a cleaner audit trail. FIFO is more granular per transaction. Each sale is tied to a specific purchase batch. Auditors can trace a single unit through purchase, hold, sale.

You operate internationally and use IFRS. IFRS allows FIFO and weighted average but bans LIFO. FIFO is the more common pick under IFRS.

Your costs are mostly stable or falling. FIFO and weighted average converge when prices are stable. If you have stable supplier pricing, FIFO is slightly more transparent.

## When weighted average is the right pick

High volume of identical SKUs at variable prices. If you buy and sell the same SKU dozens of times a month at varying prices, FIFO produces a complicated lot-by-lot ledger that is hard to manage. Weighted average gives you one number.

Commodity-style products. If you sell pallets of screws, bags of flour, or barrels of oil where individual units are not distinguishable, weighted average matches reality.

You want smoother gross margins. Weighted average smooths out the price volatility from individual purchase batches. Your monthly gross margin is more stable, which is easier to forecast.

Tax simplicity. Weighted average produces one cost per period instead of a per-batch cost trail. The tax computation is simpler.

Inflation environment. Weighted average produces a moderate COGS instead of FIFO’s “lower COGS now, higher tax now” pattern.

For most small businesses with high SKU turnover and commodity-style products, weighted average is the simpler and slightly more tax-efficient pick. For boutique retailers with low SKU count and high-value items, FIFO is cleaner.

## What about LIFO?

Last in, first out. The opposite of FIFO. You assume you sold the most recently purchased units first, leaving the older cheaper units on the balance sheet.

Why anyone uses LIFO: in an inflationary environment, LIFO maximizes COGS (using the most recent expensive prices) which minimizes taxable income. It is purely a tax minimization strategy.

Why most businesses do not: LIFO is banned under IFRS, so any company that ever wants to operate internationally or be acquired by a public company has to switch off LIFO eventually. The switch is painful and expensive. Most accountants today recommend against starting on LIFO unless you have a very specific tax-driven reason.

ERPClaw does not implement LIFO in v1. We may add it later for US-only customers who specifically want it, but it is not on the v1.x roadmap.

## How ERPClaw handles inventory costing

ERPClaw ships with both FIFO and weighted average in the core inventory module. You pick the method per item. Most items default to weighted average; you can override per SKU. See the AI inventory and costing workflows and the wider supply chain feature set for the surrounding operational flow.

Setting the costing method. When you create an item with `add-item`, the `valuation_method` parameter accepts `FIFO` or `WeightedAverage`. The default is `WeightedAverage` because it covers most small business cases.

FIFO implementation. ERPClaw maintains a per-item lot ledger. Every purchase creates a lot with its own cost. Every sale consumes the oldest lot first, posting COGS at that lot’s cost. The lot ledger lives in the `stock_ledger_entry` table, immutable, append-only, with the same audit guarantees as the GL.

Weighted average implementation. ERPClaw maintains a running per-item average cost. Every purchase recomputes the average: `(existing_qty × existing_avg + purchase_qty × purchase_cost) / (existing_qty + purchase_qty)`. Every sale uses the current average for COGS.

Switching methods. If you start on weighted average and decide to switch to FIFO, the migration requires re-creating lots from purchase history, which ERPClaw can do via a one-time `recompute-inventory-costing` action. The recompute reads all historical purchases, builds the lot ledger, and posts adjusting GL entries to reconcile. Audit-trail-friendly.

Decimal precision. All costs are stored as `Decimal`, never `float`, to avoid the sub-penny accumulation errors that plague spreadsheet-based inventory tracking.

## Common mistakes small businesses make

Mistake 1: tracking inventory in a spreadsheet at retail price. Retail price is not cost. Your spreadsheet looks like an inventory ledger but cannot produce COGS because it does not know what you paid. Fix: track cost separately or move to a real inventory system.

Mistake 2: averaging cost in your head. “I think they cost about $12 each” is not a valuation method. Pick FIFO or weighted average and let the system compute it.

Mistake 3: never doing a physical count. Books say 200 units on hand; warehouse has 178. The 22 unit gap is a real loss that must be written off. Annual physical counts (or cycle counting weekly) are a baseline control.

Mistake 4: changing methods without filing Form 3115. The IRS requires consistency in inventory methods. Switching mid-year without the form is a compliance issue. Pick at start, or file the form to switch.

Mistake 5: forgetting freight and duties. “Cost” is not just what the supplier billed for the goods. It includes inbound freight, customs duties, insurance, and any direct cost of acquisition. Most small businesses miss these and underreport COGS by 5% to 15%.

Mistake 6: ignoring shrinkage. Theft, breakage, spoilage, miscounts. Real businesses lose 1% to 3% of inventory to shrinkage. Recognize the loss as a COGS or expense entry; do not pretend the units are still on the shelf.

ERPClaw catches the first four directly via the structure of the stock_ledger_entry table. The last two are workflow choices the operator makes.

## Try it on real data

If you want to see FIFO and weighted average run against your own purchase history, ERPClaw is free and open source. Install in five minutes, import a CSV of purchases and sales, run the inventory valuation report, and see the numbers your accountant has been computing by hand.

- Install ERPClaw (5 minutes)

- Try the demo (browser, no install)

- See pricing (it is $0)

- Read about Cost of Goods Sold for Shopify stores

- See the QuickBooks alternative comparison

ERPClaw is open source, AI-native, and ships with both FIFO and weighted average in the core install. No upgrade tier, no per-user fee, no enterprise gate.

## FAQ

### Can I use different costing methods for different items?

Yes. ERPClaw lets you set the valuation method per item. Most businesses pick one default and override only for specific product lines that need the other method. For example, perishables on FIFO, commodities on weighted average.

### Once I pick a method, can I change it later?

Yes, but you have to follow the rules. The IRS requires consistency, and changing methods mid-year requires Form 3115 and a multi-year adjustment. ERPClaw can recompute the historical valuation under the new method, but talk to your accountant about the tax filing first.

### Does FIFO require lot or batch tracking?

ERPClaw’s FIFO implementation uses an internal lot ledger keyed by purchase date and cost. You do not have to track external lot numbers (the kind printed on a manufacturer’s label) unless you separately need them for compliance. The internal lot ledger is automatic.

### What about the lower of cost or market rule?

Both FIFO and weighted average value inventory at cost. If the market value drops below cost (your supplier slashed prices, or your goods became obsolete), GAAP requires you to write inventory down to market value. ERPClaw supports inventory write-downs via the `write-down-inventory` action; the GL posts the loss to the appropriate expense account.

### Does ERPClaw support LIFO?

Not in v1. LIFO is allowed in the US but banned under IFRS, and most accountants today recommend against new LIFO implementations. We may add LIFO in a later release for US-only customers with a specific tax-driven need.

### How accurate is the cost calculation?

All costs are stored as `Decimal`, not float, so there is no sub-penny accumulation error. Every cost calculation rounds explicitly using `ROUND_HALF_UP` to two decimal places. The valuation matches what an auditor would compute by hand, exactly.

### Does the inventory module run on PostgreSQL?

Yes. ERPClaw is database-agnostic via PyPika. The inventory module runs on SQLite (default) or PostgreSQL with the same code. The stock_ledger_entry table and all valuation queries work on either backend.

Install ERPClaw and run FIFO vs weighted average against your own data to see the dollar difference.

---

## Model Independence, With the Receipt
URL: https://www.erpclaw.ai/blog/model-independence-with-the-receipt/

# Model Independence, With the Receipt
ERPClaw has run in production through four different models with no module changed. Why that is possible, why most AI products cannot claim it, and what it tells a buyer.

Short answer. Production has run through Claude Opus 4.8, Kimi K2.6, Gemini 3.5 Flash, and Codex GPT-5.5, all on the OpenClaw platform, with no change to a single module. That is possible because no module was ever trained, fine-tuned, or prompt-engineered for a particular model. Swapping the model costs a configuration change. Disclosure: we build ERPClaw, so weigh the framing accordingly.

## Why this is worth stating

Every AI product says it is model-agnostic. Almost none of them can describe what would break if the model changed, which is the question that actually distinguishes the claim.

The test is not whether the product runs on more than one model in a demo. It is whether correctness depends on which one. If a product’s accuracy is materially different across models, the model is load-bearing, and the buyer is exposed to a vendor’s roadmap, pricing, and deprecation schedule in a way nobody mentioned during the sale.

## What makes the swap cheap

Three properties, and all three have to hold.

No module was trained for. There is no fine-tune, no model-specific adapter, no weights we ship. A module is a manifest and code. The model reads the manifest the way a chef reads a recipe, which is the argument we made in adding an industry is a recipe, not retraining.

The model interprets, deterministic code writes. The model works out what you meant. It never decides what the books say. Every posting goes through code that behaves identically regardless of which model triggered it, so the output of the write path does not vary with the model at all.

Invariants verify afterwards, unconditionally. Structural checks run after every ledger operation, and they do not know or care which model was in the loop. If a new model produced a wrong interpretation that led to an unbalanced book, the check fails the same way it would for any other cause. The same reasoning drives write ownership: assume the writer might be wrong, and arrange things so being wrong is survivable.

Put together: the model sits at the interpretation boundary, and everything downstream of that boundary is deterministic and checked. That is the entire reason a swap is a configuration change rather than a revalidation project.

## What the swap does not fix

Being precise about the limits, since a claim this convenient invites suspicion.

Interpretation quality does vary. A weaker model misreads ambiguous input more often. It will not produce a wrong ledger, because the deterministic layer and the invariants stand between it and the books, but it will ask more clarifying questions and get more first attempts wrong. Model independence is a claim about correctness, not about experience.

Latency and cost vary considerably. A fast, cheap model and a frontier model are not interchangeable from an operator’s point of view even when they are interchangeable from a correctness point of view.

Four models is four. It demonstrates the property rather than exhausting it. A model with a materially different tool-calling contract could require platform work, and we would say so rather than pretend the abstraction is total.

## Why most products cannot claim this

Not because their engineers are worse. Because of where the AI was added.

A system that bolted a model onto a schema designed before AI existed has to make the model responsible for correctness, since the deterministic layer underneath was never built to validate what an agent proposes. Once correctness depends on the model, changing the model means revalidating everything, and that is a project rather than a setting.

Which turns a question that sounds like procurement trivia into an architectural tell. Ask a vendor what breaks if the model changes. A specific answer means they know where their boundary is. A vague one means the model is holding more than they have admitted.

## What it means for a buyer

Three practical consequences.

You are not exposed to one vendor’s pricing. When a model gets more expensive or a cheaper one becomes good enough, that is a decision you can act on rather than absorb.

You are not exposed to deprecation. Models are retired on schedules set by other companies. A system where that is a configuration change rather than a migration is one less thing on your risk register.

And you can put a different model on different work if you want to, since nothing in the ledger layer needs to know.

## The verdict

Model independence is easy to assert and hard to arrange, and the difference between the two is where a system’s correctness lives. If it lives in the model, the claim is aspirational. If it lives in deterministic code with invariants checking afterwards, the model becomes a component, and components are replaceable by definition.

This is one of the explanations we developed for the IEEE International Conference on Information Reuse and Integration in July 2026. The receipt is the part that mattered to that audience: not that the architecture permits a swap, but that four have already happened.

Tagsai-nativearchitecturemodel-independenceerpvendor-lock-in
Share:XLinkedInHacker News

---

## MRR vs ARR in Stripe: Definitions, Formula, and Example
URL: https://www.erpclaw.ai/blog/mrr-vs-arr-stripe-explained/

# MRR vs ARR in Stripe: Definitions, Formula, and Example
Calculate MRR and ARR from Stripe subscriptions, with formula boundaries, annual plan normalization, status rules, a worked example, and the difference from recognized revenue.

Direct answer: MRR is recurring subscription value normalized to one month. In this guide, ARR means current MRR multiplied by 12. Stripe cash collected and ASC 606 recognized revenue are separate measures. Some organizations reserve ARR for contracted annual recurring revenue, so state your definition before comparing figures.

- MRR: Answers what active recurring subscription value is worth in one normalized month.

- ARR in this guide: Answers what the current monthly run rate equals when annualized as MRR × 12.

- Stripe cash collected: Answers how much cash was processed during a period, including timing effects and non-recurring items.

- Recognized revenue: Answers how much revenue belongs in a reporting period under the applicable accounting policy.

Use the ERPClaw MRR reports reference for the implementation path. Stripe also documents its Billing analytics and MRR methodology separately from its revenue recognition methodology.

A founder I know walked into a board meeting last spring with $52K
MRR on the deck. Two weeks later, in a fundraising call with the
same numbers reformatted, the figure was $61K. Same company, same
month, same Stripe account, two different numbers. He had not lied
once. He had calculated MRR two different ways, and nobody noticed
until an associate at the lead fund pulled his own data from the
Stripe API and asked which version was real.

If you are reading this, you know what MRR and ARR are. The problem
is not the definition. The problem is that Stripe gives you fifteen
plausible ways to compute each one, and the gap between “what
Dashboard shows” and “what your board wants to see” is wide enough
to drive a Series A through.

This post covers MRR vs ARR Stripe math the way you actually have to
do it: which subscriptions count, how to handle annuals, trials,
prorations, pauses, and which tools get it right.

## The simple definitions

MRR (Monthly Recurring Revenue): the sum of all active recurring
subscription revenue, normalized to one calendar month. Annual
subscriptions get divided by twelve. Quarterly by three. Weekly
multiplied by 52/12.

ARR (Annualized Run Rate): MRR multiplied by twelve.

That is it. ARR is a derived metric. If you know MRR you know ARR.
If two people compute the same MRR honestly, they get the same ARR.
The mistakes never live in the multiplication; they live in what
gets included in MRR in the first place.

Three non-obvious things implied by those definitions:

- ARR is not what you expect to collect over the next 12 months. It
is a snapshot of the current run rate annualized. Treating ARR as
a forecast is a category error.

- MRR is a point-in-time metric. “MRR for March” means “MRR as of
March 31.” It is not the cash collected in March.

- MRR only counts recurring revenue. One-time fees, setup
charges, and professional services do not belong in MRR.

Internalize those three and most arithmetic mistakes disappear.
Then Stripe enters the picture.

## Why Stripe makes MRR vs ARR Stripe math hard

Stripe is excellent at processing payments. Stripe is mediocre at
telling you what your MRR is, because Stripe sees every dollar of
recurring and non-recurring revenue through the same plumbing and
trusts you to label it correctly. The complications stack up fast.

Annual subscriptions show up as a single $1,200 charge once a
year. If you read “March charges = $1,200” and call that March MRR,
you will publish a number 12x too high every March and 0 every
other month.

Mid-cycle upgrades create proration line items. The proration
is not recurring, so it is not MRR, but it lives in the same Stripe
invoices table as the actual subscription line.

Refunds reduce cash but do not, by themselves, reduce MRR.
Whether a refund affects MRR depends on whether the underlying
subscription was cancelled.

Churn is silent in Stripe. A subscription set to
`cancel_at_period_end = true` is still active until the period ends.
Some founders book churn on the cancellation request date, others
at period end. Pick one and stick with it.

Trials create subscriptions in `trialing` status with a
non-zero plan amount but no payment. Most rigorous SaaS definitions
exclude trials; Stripe’s own dashboard sometimes includes them.

Paused subscriptions are still subscriptions. No billing, no
churn. Exclude from MRR and exclude from churn until they resume
or cancel.

Free plans are not MRR. Founders still sometimes include them
because they appear in the active customer count. Resist.

Multi-currency subscriptions need to be converted to a single
reporting currency at a consistent FX rate. Pick spot-at-month-end
or trailing-90-day average. Document the choice.

Past-due (delinquent) subscriptions are nominally active but
the payment failed. Industry convention: include for a 7 to 30 day
grace window, then exclude. Document your policy.

Each of these has a correct answer. Each correct answer is a policy
choice, and most founders make those choices implicitly, then
unknowingly make a different choice next quarter. Hence the $52K to
$61K spread.

## The mistakes most founders make

In rough order of frequency:

1. Using cash collected as MRR. “We did $80K in March, so MRR
is $80K.” No. Cash includes annual prepayments, one-time fees,
late payments, and proration top-ups. Cash is cash. MRR is the
normalized recurring run rate as of a date.

2. Including trials. If 40 of your 200 “paying” customers are
on day 3 of a 14-day trial, MRR is $15,840 not $19,800. Including
trials hides churn at the trial-to-paid boundary, exactly where
most SaaS leakage happens.

3. Forgetting to divide annuals by 12. “We closed three annual
$24K deals, so MRR went up $72K.” MRR went up $6K. ARR went up
$72K. Conflating these is the single most common error in pitch
decks.

4. Ignoring proration on upgrades. A customer going from $100
to $200 mid-month increases the recurring run rate by $100, not
$150. The $50 proration is one-time.

5. Counting paused subscriptions as active. Stripe shows them
as `paused` not `cancelled`. If your sum-of-active query does not
filter on status, paused subs sneak in.

6. Including non-recurring add-ons. Setup fees, training fees,
implementation fees. The “R” in MRR is doing actual work.

7. Trusting Stripe Dashboard’s MRR blindly. Dashboard rounds
per-subscription, includes some trial states, and has changed
methodology at least twice since 2020. Use it as a sanity check,
not a reporting source.

8. Mixing booked and recognized revenue. Booked MRR is the
normalized run rate. Recognized revenue (the GAAP number) follows
ASC 606 and ratably amortizes annual contracts across 12 months.
For MRR you want booked. We covered the recognition side in the
Stripe ASC 606 revenue recognition guide.

9. Inconsistent treatment of discounts. A $100 plan with a
permanent 50% off coupon is $50 MRR. A $100 plan with a one-month
introductory discount is $100 MRR. Most spreadsheets get this
wrong half the time.

Avoid those nine and your number will be defensible.

## The right way: a procedure that survives an investor’s pull

A procedure that produces a single, reproducible MRR number on any
given date:

- Pull all subscriptions with status in (`active`, `past_due`)
as of your reporting date. Exclude `trialing`, `paused`,
`incomplete`, `incomplete_expired`, `canceled`. Apply your
`past_due` grace policy.

- For each subscription, sum its line items. Each line is
`unit_amount` times `quantity` times any active discount.

- Normalize to monthly. Yearly /12, quarterly /3, monthly x1,
weekly x52/12, daily x365/12.

- Subtract permanent discounts. Apply percent-off coupons to
the line amount before summing. Skip one-time and limited-time
discounts.

- Convert to your reporting currency at a consistent FX rate.

- Sum across all subscriptions. That is your MRR.

- Multiply by 12. That is your ARR.

That procedure is what
ERPClaw’s MRR engine does, and it is
roughly what Baremetrics, ChartMogul, and ProfitWell do too. The
point is the procedure, not the tool. You should be able to point
at the line in your code or spreadsheet that implements each step.

If your investor pulls the Stripe API and runs the same seven steps
with the same policy choices, they should get the same number you
do. If they do not, one of you is making a different policy choice,
and you should know which.

## A worked example: Acme SaaS

Let’s actually compute it.

Acme SaaS, as of March 31, 2026:

- 47 monthly subscriptions on the Pro plan at $99/month

- 12 annual subscriptions on the Pro plan at $1,200/year

- 8 customers on a 14-day trial (excluded)

- 3 customers paused (excluded)

- 2 customers in `past_due` for 4 days (included; within grace)

- 1 customer with a permanent 20% off coupon on the $99 plan
(counted as $79.20)

- 5 customers on the free Hobby plan (excluded; not recurring
revenue)

- $4,500 in one-time setup fees collected in March (excluded; not
recurring)

The math:

| Bucket
| Count
| Per-sub MRR
| Subtotal

| Monthly Pro at $99 (no discount)
| 46
| $99.00
| $4,554.00

| Monthly Pro at $99 with 20% off
| 1
| $79.20
| $79.20

| Annual Pro at $1,200/yr
| 12
| $100.00
| $1,200.00

| Past-due monthly (within grace)
| 2
| $99.00
| (already in 46 above)

| MRR total
|
|
| $5,833.20

ARR = $5,833.20 x 12 = $69,998.40

Total active subs in MRR: 47 monthly + 12 annual = 59. Trials,
paused, free, and one-time fees excluded.

The deck-friendly numbers:

- MRR: $5,833 (or to the cent if your audience is finance)

- ARR: $70,000 (round to nearest thousand for investor decks)

What you do not report: cash collected (includes prepayments
and one-time fees), customer count (includes trials, paused, free),
or “what ARR would be if trials convert” (speculation, not a metric).

What makes the number defensible: every line is traceable to a
Stripe object, every exclusion has a documented reason, and the
math is a sum, not an estimate. An investor asking “where does the
$5,833 come from” can be handed the per-customer breakdown and walk
away satisfied in five minutes.

## Tools that calculate MRR for you

If you do not want to maintain the procedure above by hand, here
are the main options. Honest one-line assessment of each.

- Stripe Dashboard’s built-in MRR. Free, on by default with
Stripe Billing. Approximately right, but rounds per-subscription
and sometimes includes trials. Sanity check, not source of truth.

- ChartMogul. From $59/month on paid
tiers (as of June 2026). Mature, well-documented, used
by a lot of Series A SaaS companies. Strong cohort retention.

- Baremetrics. From $75/month on paid
tiers (as of June 2026). Cleaner UI than ChartMogul,
with a public-dashboard feature some founders use as a trust
signal. Their methodology docs are worth reading even if you do
not subscribe.

- ProfitWell (now Paddle Metrics). Free tier exists. Strong on
retention analytics and industry benchmarking. The free tier is
genuinely useful under $10M ARR.

- Stripe Sigma. Pay-per-query SQL on your Stripe data. Powerful
if you have a data person; useless if you do not.

- ERPClaw. Free, open source, AI-native, self-hosted on your own infrastructure.
Computes MRR, ARR, churn, expansion, contraction, and cohort retention
from Stripe subscription data, and posts the underlying journal entries
to a real GL. If you only need MRR reports, ERPClaw is overkill.
If you also want ASC 606 revenue recognition, three-layer payout
reconciliation, and an actual general ledger, it is the only free
AI-native option that does all three. See the
Stripe MRR reports doc and the
Rillet comparison for the head-to-head.

The right tool is the one that makes your reporting cadence faster
without changing your number. If you switch tools and your MRR
moves more than a couple of percent, the tool was wrong, your old
number was wrong, or both. Restate the prior periods and document the change.

## What MRR alone hides

Here is the part most founders do not internalize until they get
asked about it on a partner call.

Cohort retention is the metric MRR cannot tell you. Two
companies can both have $100K MRR. Company A keeps 95% of each
cohort a year out. Company B keeps 60%. Identical headline,
completely different businesses.
OpenView’s annual SaaS benchmark report
publishes cohort retention curves by ARR band; comparing your curve
to the median in your band is one of the fastest ways to know if
you have a real product or a leaky bucket.

The way to read a cohort retention table:

| Cohort
| M0
| M3
| M6
| M9
| M12

| Jan 2026
| $5,000
| $4,800
| $4,650
| $4,500
| $4,000

| Feb 2026
| $7,000
| $6,700
| $6,500
| $6,300
| Not available

| Mar 2026
| $9,500
| $9,100
| $8,900
| Not available
| Not available

Each row is a single signup cohort. M0 is MRR at signup, M12 is
what is left a year later. Decay across the row is churn. Flat or
up means net negative churn (expansion offsets churn): a special
business. If headline MRR grows but every cohort row collapses
from $5,000 to $4,000 in 12 months, you are buying growth that
leaks out the back. Investors spot it within minutes.

## Logo churn vs revenue churn: which to trust

Two definitions, both legitimate, both report-worthy:

Logo churn = (customers who cancelled in period) / (customers
active at start of period). What percentage of your customers left,
counted by company.

Revenue churn = (sum of MRR lost to cancellation in period) /
(MRR at start of period). What percentage of your revenue left,
counted by dollars.

These two diverge when customer size varies. Say you have 100
customers paying $100 and 1 customer paying $10,000: your MRR is
$20,000. If you lose the one big customer, your logo churn is 1%
and your revenue churn is 50%. Same event, two stories.

Which one to trust depends on what you are trying to learn:

- Product-market fit: logo churn. If 30% of new signups leave
within 90 days regardless of size, the product is the problem.

- Revenue durability: revenue churn. If a few enterprise deals
carry your business, watch the dollars.

- Investor reporting: report both. Smart investors ask if you
only report one.

Cohort retention plus split logo/revenue churn tells you the actual
health of a SaaS business in ninety seconds. MRR alone does not.

## Closing

MRR vs ARR Stripe math is not arithmetic. It is policy. The math is
trivial; the policy choices about what to include and exclude are
where two honest founders end up with two different numbers.

If you take three things from this post:

- Pick a procedure, write it down, and apply it the same way every
month. The number being defensible matters more than the number
being optimized.

- ARR is MRR x 12. Anything else dressed up as ARR is a forecast,
not a metric. Investors know the difference.

- MRR alone hides everything that matters. Pair it with cohort
retention and split logo/revenue churn or your dashboard is
lying by omission.

If you want a tool that computes this from your Stripe data without
sending it through someone else’s servers, ERPClaw is one free
option. The Stripe integration page, the
SaaS landing page, and the pricing cover
what is and is not included.
ChartMogul, Baremetrics, and ProfitWell
are good paid alternatives. The right choice is the one that makes
your number boring to compute.

Boring numbers are the ones investors trust.

## Frequently asked questions

### What is the difference between MRR and ARR Stripe-side?

Stripe stores subscriptions with billing intervals and amounts, not
either metric directly. MRR is computed by normalizing every active
subscription to a monthly rate and summing. ARR is that sum
multiplied by 12. Stripe Dashboard shows its own MRR computation,
which is approximate and best treated as a sanity check.

### How do I calculate MRR from Stripe with annual subscriptions?

Pull every active annual subscription, divide each annual amount
by 12, and add to your monthly subscription total. A $1,200/year
customer contributes $100 to MRR. Do not include the $1,200 cash
payment in the month it was charged.

### Should I include trials in MRR?

No. Trials have not generated cash and may never convert. Including
them inflates MRR and hides churn at the trial-to-paid boundary.
Track trial counts and conversion rates separately as a leading
indicator.

### What about past-due subscriptions?

Industry convention is to include for a 7 to 30 day grace window
then exclude. The grace window gives dunning time to recover the
payment. Pick a window, document it, do not change it without
restating.

### Does Stripe Dashboard’s MRR match what my investors will compute?

Probably not exactly. Expect a small gap, typically a few percent,
when an investor pulls the API and runs their own computation. The gap reflects different
policy choices. The question is whether your reported number ties
to a procedure you can defend.

### What is annualized run rate Stripe-side, and is it the same as ARR?

For pure-Stripe SaaS, yes: current MRR multiplied by 12. Some tools
distinguish “ARR” (contractually committed annual revenue) from
“run rate” (current MRR annualized) for companies mixing
month-to-month and annual contracts. For most Stripe-only reporting,
the two collapse to the same number.

### Should I report booked MRR or cash MRR to investors?

Booked MRR. Cash MRR is useful for runway modeling but is not the
standard SaaS metric. Most decks report booked; if you report cash,
label it.

### How does ASC 606 revenue recognition relate to MRR?

Different metrics. MRR is a normalized run rate snapshot. ASC 606
recognized revenue is a GAAP measure that ratably amortizes annual
contracts across the service period. They will not match
month-to-month and they should not. Use MRR for SaaS metrics, use
recognized revenue for financial statements. The
ASC 606 guide
covers the recognition side in depth.

---

## The NetSuite Alternative for Teams Who Saw the Quote
URL: https://www.erpclaw.ai/blog/netsuite-alternative-mid-market/

# The NetSuite Alternative for Teams Who Saw the Quote
NetSuite is the default mid-market ERP and the quote shows it. ERPClaw is the AI-native open-source alternative at roughly 100x less. An honest comparison.

You got the NetSuite quote.

It probably looked something like this: roughly $25,000 a year in license, about $150,000 for implementation through the partner Oracle steered you toward, and a nine month timeline before you go live. Probably more. Your COO read it, your CFO read it, and now somebody on the team is searching “NetSuite alternative” at 11pm on a Sunday.

This post is for that person.

I am not going to tell you NetSuite is bad. NetSuite is the closest thing the mid-market has to a real ERP standard. Oracle has spent two decades building it, more than 43,000 organizations run on it (as of June 2026), and the SuiteApp ecosystem covers verticals most products do not even acknowledge. If you have $200,000 in budget and a nine month runway, NetSuite is a reasonable choice. I used to roll out SAP for a living and I will say the same about SAP. And if the quote on your desk is for SAP Business One rather than NetSuite, the SAP Business One alternatives in 2026 rundown walks that exact path.

But “reasonable choice for the buyer who can afford it” is a different sentence from “right tool for a ten million to one hundred million revenue mid-market company in 2026.” That second question is the one this post is about.

ERPClaw is a free, open source (open source), AI-native ERP that covers the same primitives NetSuite does: accounting, inventory, manufacturing, HR, payroll, CRM, projects, billing, multi-entity, multi-currency. All included modules, one shared database, zero per user fees. Implementation is five minutes instead of nine months. The pitch is not “we replace NetSuite at the Fortune 500 level.” We do not. The pitch is “we cover roughly the same primitives at the mid-market entry point for about 100x less, with an AI-native architecture NetSuite cannot retrofit onto a 1998 codebase.”

Math, gaps, and the honest cases where you should still pick NetSuite below. If your shortlist also includes the open-source ERPs, the ERPNext alternatives in 2026 roundup covers that side of the field.

## What NetSuite genuinely does well

I want to be careful here because most “NetSuite alternative” posts open with grievances and that is not the honest take. NetSuite earned its position.

Real depth across modules. NetSuite’s accounting is audit grade and has been since before most of its competitors existed. Inventory handles serialization, lot tracking, multi-warehouse, kitting, drop ship, intercompany transfers. Manufacturing covers BOM, routing, work orders, and MRP. The GL has been hardened by more than 43,000 customers across 20 years.

Multi-entity and multi-currency consolidation. If you are a holding company with eight subsidiaries in five countries, NetSuite OneWorld handles intercompany eliminations, automated FX consolidation, and statutory reporting per jurisdiction. This is the genuinely hard part of accounting.

The SuiteApp ecosystem. Avalara for tax, Celigo for integration, FloQast for close management, hundreds of vertical apps. If you have a weird requirement, somebody built a SuiteApp for it.

Premium support and Oracle’s backing. You get an account manager, a partner who has done your industry 50 times, and a contractually enforced SLA. Oracle is not going anywhere. The product will exist in 10 years. For some buyers that stability is worth the price tag.

The case for ERPClaw is not that any of those are missing or wrong. The case is that for a mid-market revenue company, the price you pay for that depth is wildly disproportionate to what you actually use, and the AI-native architecture flips the cost curve in a way Oracle’s pricing model cannot survive.

## Where ERPClaw matches the primitives

NetSuite is a sprawling, deep product. ERPClaw covers the same back-office surface. The overlap is the part most people do not realize exists. ERPClaw includes, in the core install, sharing one database, with no upcharge per user or module:

- Accounting: full double-entry GL, a US chart of accounts out of the box, AR/AP aging, trial balance, P&L, balance sheet, cash flow, multi-company, multi-currency, period close. Submitted GL entries are immutable. Every cross-table write is one transaction. Every posting passes a full validation pipeline before it touches the books.

- Inventory: items, warehouses, batches, serial numbers, reorder levels, FIFO and weighted average costing, transfer orders, stock revaluation.

- Manufacturing: bill of materials, routing, work orders, production planning, MRP.

- HR and Payroll: employees, time off, expense claims, salary structures, FICA, federal and 50 state withholding, W-2 generation. In the core, not an add-on.

- CRM: leads, opportunities, pipelines, contacts. Enough for a 50 person sales team.

- Projects and Billing: project P&L, time entries, recurring invoices, usage-based billing, ASC 606 revenue recognition.

- Stripe deep integration: all core Stripe actions, three-layer payout reconciliation, ASC 606 engine, Connect platform fees.

- Shopify deep integration: all core Shopify actions, OAuth pairing, GDPR webhooks.

- Multi-entity: multiple companies in one install, intercompany journal entries, basic consolidation.

- Industry verticals: retail, restaurant, healthcare, legal, nonprofit, education, real estate, agriculture, automotive, food, hospitality, construction, fleet, logistics, and more.

Total those bullets as separate SaaS subscriptions and a 50 person company is past $4,000 a month in software rent. ERPClaw is $0 forever (open source), self-hosted on your own infrastructure, scales to PostgreSQL via PyPika when you outgrow SQLite.

The structural point: when accounting, inventory, payroll, manufacturing, and CRM all live in one database, your AR aging knows the customer’s CRM history, your inventory valuation knows the manufacturing routing, and your project P&L knows the actual labor hours from HR without a sync job. The connections between these things are not “supported integrations,” they are inherent. Nothing to break at 11pm.

## Where NetSuite still wins

I am going to be honest about this because the alternative would make me look credible to nobody.

Multi-currency consolidation at scale. ERPClaw handles multi-currency in the core. What we do not yet match is OneWorld’s automated intercompany eliminations across eight legal entities in five countries with statutory IFRS reporting per jurisdiction. We will get there. Not today.

The SuiteApp ecosystem. If your business depends on a specific SuiteApp, it does not exist for ERPClaw and may not for years. Our answer is “you can write it in a weekend with Claude Code because the codebase is open,” but for some buyers that is not the right answer.

Premium support SLAs. NetSuite contractually guarantees a response time. ERPClaw is a small co-founder team plus a Discord and `support@avansaber.com`. If your audit committee requires that contractual posture, you need NetSuite or Sage Intacct.

Established sales motion and Oracle’s backing. Quarterly business reviews with a named account executive, a customer success manager, a publicly traded vendor with an enterprise-scale market cap. If procurement requires those, we cannot satisfy that.

For a Fortune 500 buyer, those gaps are dealbreakers. For a typical mid-market buyer, they are usually theoretical: nice on paper, not the reason work gets done. The honest question is which group you are in.

## The AI-native cost argument

Here is the part of the post that matters.

NetSuite cannot be cheap. Oracle has 160,000 employees. NetSuite has a sales force, a partner channel that needs implementation revenue to stay healthy, a customer success org, a support organization, and shareholders. The pricing model is downstream of the cost structure. A typical mid-market NetSuite deal (partner-quoted ranges; NetSuite does not publish list pricing; contact Oracle for a current quote):

- License: $25,000 to $50,000 per year for 20 to 50 users

- Implementation through a partner: $100,000 to $300,000 one time

- Annual maintenance and consulting: $30,000 to $80,000 a year

- Avalara, Celigo, and a couple of SuiteApps: another $20,000 to $40,000 a year

- A NetSuite admin on payroll: $90,000 to $130,000 fully loaded

Five year total cost of ownership for a 50 person mid-market company on NetSuite: somewhere between $750,000 and $1.5M. Not a typo.

ERPClaw is $0 in license, self-hosted on your own infrastructure, and the implementation is `clawhub install erpclaw` plus telling it about your business in plain English. Five year total cost of ownership: $0 in software fees. Whatever you pay your existing accountant or controller to run the books, you were going to pay anyway.

That is roughly 100x cheaper at the mid-market entry point. Not 2x. Not 10x. Two orders of magnitude.

This is possible for one structural reason that NetSuite cannot replicate without a full rewrite. ERPClaw is AI-native, not AI-decorated. The full argument is in AI-decorated vs AI-native software. The short version: NetSuite was built between 1998 and 2024 by humans writing SuiteScript. Adding AI means bolting NetSuite AI onto an existing codebase, sold as an upsell, with the underlying cost base unchanged. ERPClaw was specced first and built by Claude Code, so the marginal cost of building module 49 or vertical 15 is hours, not quarters. It can be free because the cost to produce it dropped by an order of magnitude.

The AI-native posture is also why ERPClaw is conversational. You talk to it through Telegram, WhatsApp, the web dashboard, or any OpenClaw client. For example: “Create a sales order for Acme, 500 units of widget A at $24, due in 30 days.” It writes the order, posts the inventory commitment, schedules the GL entries on shipment. Native interface, not a chatbot wrapper around forms. NetSuite is, structurally, a forms application with an AI sidebar.

ERPClaw is also database-agnostic. Default install is SQLite (one file, you own it). When you outgrow that, swap to PostgreSQL via PyPika without rewriting application code. NetSuite runs on Oracle. You do not get to pick.

## Side by side

Full comparison page is at /compare/netsuite. Pulling the table here.

| Feature
| ERPClaw
| NetSuite

| Annual license
| $0 (open source)
| $25K to $50K

| Implementation
| $0 (5 min install)
| $100K to $300K one time

| Time to live
| 5 minutes
| 6 to 18 months

| Modules included
| All included, always
| Tier-gated

| Multi-entity
| Included
| OneWorld upcharge

| Multi-currency
| Included
| OneWorld upcharge

| Industry verticals
| Native
| SuiteApp marketplace

| AI capabilities
| AI-native (every action)
| AI add-on (NetSuite AI)

| Customization
| Spec-first (any LLM regenerates)
| SuiteScript (proprietary JS)

| Database
| SQLite or PostgreSQL via PyPika
| Oracle (locked)

| Self-hosted
| Yes
| No (Oracle cloud)

| Open source
| Yes (open source)
| Proprietary

| Vendor lock-in
| None (cp data.sqlite anywhere)
| Significant

| Support model
| Community + founder direct
| Contractual SLA

| Customer count
| Early adopters
| 43,000+

Two rows deserve a second look.

Customization. NetSuite customization happens through SuiteScript, a proprietary JavaScript dialect tied to NetSuite’s runtime. Every SuiteScript developer you hire is a NetSuite-specific hire and the customizations you build are bound to the platform. ERPClaw customization happens through specs that any LLM can re-implement. If you describe a new module in plain English and feed it to Claude or Gemini or any future model, you get working code in the same shape as the rest of the codebase. The customization is portable across the AI tooling you already use.

Database. NetSuite gives you a proprietary cloud database with limited export. ERPClaw gives you a SQLite file you can literally copy to a USB stick, or a PostgreSQL backend you control. If you want your data, you have your data.

## When NetSuite is the right call

I promised honest. Here are the cases where I would tell you to write the check to Oracle and not even bother with ERPClaw.

You have the budget and the timeline. If your company has a substantial budget committed and a board mandate to be on a tier-one vendor, NetSuite is what that money buys. The implementation partner gets you live faster than a self-serve install ever will.

You depend on a specific SuiteApp. If your business runs on Avalara, Celigo, or a vertical SuiteApp you cannot replicate, those investments stay and so does NetSuite.

You need a contractual support SLA or a publicly traded vendor. Audit committees, board oversight, and PE owners often require enforceable SLAs and a vendor with a market cap. ERPClaw cannot match a contractual 4 hour P1 response today.

You are doing serious global consolidation. Eight legal entities across five countries with statutory IFRS reporting per jurisdiction is what OneWorld was built for. The consolidation depth at OneWorld scale is genuinely a gap for us today.

If none of those apply, ERPClaw is worth a weekend of evaluation before you sign the NetSuite SOW.

## A note on NetSuite migrations

If you are already on NetSuite, the honest news first: NetSuite migrations are famously hard. Not because NetSuite locks you in maliciously, but because the data model is deep and SuiteScript customizations have to be re-implemented somewhere. The standard estimate for migrating off NetSuite is six to twelve months at a budget similar to the original implementation.

ERPClaw is a viable target if your business has simplified since the original NetSuite install (most do, post-acquisition or post-divestiture), but I will not pretend it is a weekend job. Realistic path: pick a quarter end as cutover, run parallel for two months minimum, validate trial balances to the cent, then cut. We have a migration approach in /docs.

If you are pre-NetSuite, this is much easier. Install ERPClaw in five minutes, run it for 90 days alongside your spreadsheets, decide.

## FAQ

Is ERPClaw really mid-market ready, or just for solo founders?

Both. The same install scales from one user to roughly 200 on SQLite, then swaps to PostgreSQL via PyPika for larger deployments. Same engine, same validation, same modules. Early adopters in the 25 to 75 employee range run production. We do not yet have a Fortune 500 reference customer and are not pretending to.

How does AI-native actually help my finance team?

Three places. The interface: For example, your AP clerk says “post the invoice from Acme for $4,200, due net 30, code to legal expense” instead of clicking through six screens. The build: when you need a new report or a field, you describe it and it ships in hours. The integrations: any new SaaS tool can be wired in by an LLM reading the source, not by a high-cost Celigo consultant.

What happens if I outgrow SQLite?

Swap to PostgreSQL. The application uses PyPika so the same code targets either backend. SQLite handles a surprising amount of throughput, but at thousands of transactions per minute, Postgres is the right move and the migration is a config change.

Can my CFO get audit-grade financial statements out of this?

Yes. The GL is immutable, every cross-table write is one transaction, every posting passes a full validation pipeline. Trial balance, P&L, balance sheet, cash flow, AR/AP aging are standard reports. Auditors who have looked at it have approved.

Does ERPClaw match NetSuite OneWorld on multi-currency?

For the basics, yes: any ISO currency, multiple functional currencies, consolidation in your reporting currency, manual or daily FX feeds. For OneWorld’s statutory per-country reporting, automated intercompany elimination journals, IFRS to US GAAP reconciliation, we are not at parity. If those are blockers, NetSuite is the right call.

What if ERPClaw the company disappears?

You still have the software. open source license, your code, your data, your SQLite file. The repo at github.com/avansaber/erpclaw is forkable by anyone. Compare to a cloud-only ERP that changes pricing or sunsets a tier on you.

How does this compare to Sage Intacct?

Similar tradeoffs. Sage Intacct is the second-place mid-market ERP. See /compare/sage-intacct. Structural argument is identical: AI-native, open source, free, covers the primitives, gaps in deep multi-entity and ecosystem.

## Try it before you sign the SOW

The NetSuite sales cycle is six to nine months. Use the first weekend of that cycle to install ERPClaw and see if it gets you to 80% of where you need to be.

`clawhub install erpclaw`
That puts the core ERP (full accounting, inventory, HR, payroll) on your server. Then talk to it: “I run a 60 person manufacturing company in Ohio, set me up.” It picks the relevant industry vertical, installs the matching modules, generates the chart of accounts, and is ready for transactions.

If you want to read first:

- /features for the full module breakdown

- /pricing (still free)

- /compare/netsuite for the side by side

- /compare/sage-intacct if Sage is also on your shortlist

- /docs for install and migration approach

- /blog/ai-decorated-vs-ai-native-software for the architecture argument

Repo: github.com/avansaber/erpclaw. Star it if you want to follow along, we ship weekly. Email `support@avansaber.com` if you want a 30 minute call before you commit. NetSuite product page is at netsuite.com if you want to compare directly.

NetSuite is the right software if you have the budget, the timeline, and the ecosystem requirements that justify the price. ERPClaw is the right software if you got the quote, did the math, and realized you want the same primitives at roughly 1% of the cost, with an AI-native architecture that actually fits how your team works in 2026.

If your CFO is staring at a $200,000 line item for software you have not even installed yet, that is the moment to spend a weekend installing this one instead.

---

## NetSuite Alternatives in 2026: The Honest, Open-Source Take
URL: https://www.erpclaw.ai/blog/netsuite-alternatives-2026/

# NetSuite Alternatives in 2026: The Honest, Open-Source Take
A vendor-neutral guide to the real NetSuite alternatives in 2026, plus a simple test for telling genuine AI-native ERP from AI bolted onto old software.

Short answer. If you searched “NetSuite alternatives,” you fall into one of three groups. You want the same depth for less money (look at Sage Intacct or Acumatica). You want finance software built around AI from the start (look at the AI-native newcomers like Rillet or DualEntry, and at ERPClaw). Or you want to own your software and data outright (open source: ERPNext, Odoo, or ERPClaw). This guide covers all three honestly, and gives you a simple test to cut through the “AI-native” marketing every vendor now uses.

NetSuite is the default mid-market ERP for a reason. Oracle has spent two decades on it, more than 43,000 organizations run on it (as of June 2026), and the partner ecosystem covers verticals most products ignore. The reason you are reading this is almost always the quote: license in the $25,000 to $50,000 a year range, plus a six-figure implementation, plus a long runway before go-live. For a company doing ten to a hundred million in revenue, that math is worth a second look.

So here are the real alternatives, grouped by what you are actually trying to do.

## The test: genuine AI-native, or AI bolted on?

Every ERP now says “AI.” Most of it is a summarize button on a system designed in 1998. Before you shortlist anything, run three questions:

- Does the AI do the work, or describe it? A summary of your dashboard is not the same as an agent that posts the journal entries end to end. Ask the vendor to show an action being completed, not narrated.

- Was the system built around AI, or retrofitted? Adding a model to a twenty-year-old codebase keeps the old cost base and the old workflows. Software specced around AI from the first line behaves differently.

- Can you talk to it to get something done? “Create a sales order for Acme, 500 units at $24, net 30” should produce the order, the inventory commitment, and the scheduled posting. If the only interface is forms with a chat sidebar, that is AI-decorated, not AI-native. The full argument is in AI-decorated vs AI-native software.

Hold each option below to that test.

## Group 1: the traditional mid-market options

These play NetSuite’s game on similar terms. Solid, established, paid, and AI added on rather than built in.

- Sage Intacct. The usual second choice for finance-heavy companies. Strong core accounting and multi-entity. AI features are an add-on layer, not the foundation.

- Acumatica. Cloud ERP known for user-based pricing that does not punish you for adding seats. Broad functionality, AI bolted on.

- Microsoft Dynamics 365 Business Central. The right call if you already live in the Microsoft stack. Copilot is the AI story, added to an existing product.

- SAP Business One. Aimed at manufacturing and distribution. Capable, older architecture, AI as a side feature. (Having rolled out SAP, I can say it earns its keep for the right buyer and overserves everyone else.)

Pick one of these if you want NetSuite-class depth and a named vendor, and AI is a nice-to-have rather than the point.

## Group 2: the AI-native newcomers

This is the genuinely new category: finance systems specced around automation from the start. Most are venture-funded, closed source, and paid.

- Rillet. An AI-native general ledger built for mid-market finance, designed by accountants. Credible on the automation story. Closed and paid.

- DualEntry. AI-native ERP aimed at finance teams scaling toward IPO. Also closed and paid.

- Campfire and similar. A growing set of well-funded entrants in the same space.

These pass the AI-native test. The trade-off is the usual one: you rent the software, the source is closed, and your data lives in someone else’s cloud on their pricing roadmap.

## Group 3: open source (own your software)

If vendor lock-in is the thing keeping you up, this is your group.

- ERPNext and Odoo. The established open-source ERPs. Broad and genuinely yours to host. The catch against the test above: AI arrives through plugins and add-ons, so they are open source but AI-decorated, not AI-native.

- ERPClaw. This is the one we build, so weigh it accordingly. It is the option that is open source and AI-native and free: the agent posts double-entry journals end to end rather than summarizing a screen, it covers the same back-office functions NetSuite does (accounting, inventory, manufacturing, HR, payroll, CRM, projects, billing, multi-entity, multi-currency) in one shared database, and it installs in minutes. It is $0 forever, self-hosted on your own infrastructure. The honest gaps: deep global consolidation at the scale of NetSuite OneWorld, a large third-party app marketplace, and a contractual support SLA. If those are dealbreakers, a Group 1 vendor fits better, and we say so.

## Side by side

| Option
| AI
| Open source
| Self-host
| Price posture
| Best fit

| NetSuite
| Add-on
| No
| No
| Six-figure
| Funded, ecosystem-dependent

| Sage Intacct
| Add-on
| No
| No
| Paid
| Finance-heavy mid-market

| Acumatica
| Add-on
| No
| No
| Paid
| Seat-growth companies

| Dynamics 365 BC
| Add-on
| No
| No
| Paid
| Microsoft shops

| SAP Business One
| Add-on
| No
| No
| Paid
| Manufacturing, distribution

| Rillet
| Native
| No
| No
| Paid
| AI-first mid-market finance

| DualEntry
| Native
| No
| No
| Paid
| Finance teams scaling to IPO

| ERPNext / Odoo
| Plugin
| Yes
| Yes
| Free or paid tiers
| Open-source traditionalists

| ERPClaw
| Native
| Yes
| Yes
| $0 forever
| Open-source, AI-native, cost-led

The point of the table is not that ERPClaw wins every row. It is that the bottom-left of the matrix, open source and AI-native at the same time, is nearly empty. The commercial roundups skip it because there is no affiliate revenue in a free tool. That is the gap this guide exists to fill.

## Which group are you in?

- You have the budget and the ecosystem needs. Stay with NetSuite, or look at Sage Intacct. The money buys depth and a partner who has done your industry many times.

- You want AI to actually run finance, and you will pay for it. Shortlist Rillet or DualEntry, and put ERPClaw next to them so you have a free baseline to measure against.

- You want to own the software and stop renting. ERPNext, Odoo, or ERPClaw. If you also want the AI to do the work and not just describe it, ERPClaw is the one that is both.

## FAQ

What is the cheapest real alternative to NetSuite?

Open source is the floor. ERPNext and Odoo have free community editions; ERPClaw is $0 forever and self-hosted, with AI built in rather than added. “Cheapest” only matters if it also covers your back office, so check the functions you actually use before you compare prices.

Are the AI-native ERPs actually better, or is it marketing?

Both exist. Use the three-question test above: does the AI complete actions, was the system built around AI, can you talk to it to get work done. Rillet, DualEntry, and ERPClaw pass it. A summary button on a legacy product does not.

Can an open-source ERP really replace NetSuite?

For many mid-market companies, yes, for the back-office primitives. Where it does not yet match NetSuite is deep multi-country statutory consolidation and a large third-party app marketplace. If you depend on those, keep NetSuite.

Is ERPClaw mid-market ready?

It runs production for companies in the 25 to 75 employee range and scales further when you move it from SQLite to PostgreSQL. There is no Fortune 500 reference customer yet, and we do not claim one. See the NetSuite comparison for the mid-market buyer for the detailed case.

How hard is it to leave NetSuite?

Migrating off NetSuite is a real project, usually a quarter or two, because the data model is deep and customizations have to be rebuilt. Run the new system in parallel through a period close, validate the trial balance to the cent, then cut over.

## Where to go next

- ERPClaw vs NetSuite, the detailed side by side

- The NetSuite alternative for the mid-market buyer who saw the quote

- Sage Intacct alternative on a $0 budget

- The 5 AI-native ERPs that actually earn the label

If you want to try the open-source AI-native option before you sign anything, install ERPClaw in a few minutes and run it alongside your current books for a period. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

Tagsnetsuitealternativesai-nativeerpopen-sourcecomparison
Share:XLinkedInHacker News

---

## Odoo Alternative: When an AI-Native, Open Source ERP Fits
URL: https://www.erpclaw.ai/blog/odoo-alternative-ai-native-erp/

# Odoo Alternative: When an AI-Native, Open Source ERP Fits
Evaluating Odoo alternatives? Compare tradeoffs in price, AI architecture, databases, implementation, manufacturing depth, and migration readiness.

You started looking at Odoo because the QuickBooks bill went up again, the inventory module you bolted on stopped syncing, and somebody on Reddit said “Odoo is open source, just install it.”

Then you opened pricing.odoo.com.

Odoo Community is technically free. Odoo Online and Odoo Enterprise are not. The pricing page shows “One App Free” up top and then quietly walks you up to $31.10 per user per month for the Standard plan (as of June 2026) and $61.00 per user per month for Custom, both billed annually. The Custom plan shows a first-year discount to $49.00, then snaps to $61.00 in year two. A 10 person company on Custom pays $5,880 in year one, $7,320 every year after that. That is before any partner implementation fee, before any Odoo Studio customization, and before the inevitable third-party module subscriptions. The “free open source ERP” turned out to be a tier that locks you out of Odoo Studio, the mobile app, the sales support, and most of the apps people actually want.

This post is for you if you Googled “Odoo alternative” because the math stopped feeling like open source.

This guide owns the buyer fit decision. Use the ERPClaw vs Odoo comparison for the feature grid, the Community vs Enterprise guide for edition choices, the pricing analysis for the dated cost snapshot, and the migration playbook for export and cutover.

ERPClaw is a free, open-source (open source), AI-native ERP. The full operations surface in one shared database. No per-user fee. No paid tier. No Enterprise edition with the features you actually need locked behind it. Self-hosted on your own machine or any cloud server you want. Built in 2026 with AI-native architecture from line one, not an ERP from 2005 with AI bolted onto the side.

This is an honest comparison. Odoo is a real product with real strengths. I will name them. The case for the Odoo alternative is structural, not “Odoo is bad.”

## Quick verdict

If you want a free open-source ERP and you are running a 1 to 50 person business, ERPClaw is the cleaner pick because there is no paid tier, no enterprise lock-in, no per-user fee, and the AI-native architecture means the product gets smarter with every release instead of trying to retrofit AI onto a 20 year old codebase.

If you are a 200 person manufacturing company in Belgium with a partner already implementing Odoo Enterprise, you are probably better off finishing that project. Odoo’s manufacturing depth at the top tier is real and ERPClaw is younger.

For everyone in between, the structural points below matter.

## What Odoo genuinely does well

I will not pretend Odoo is a weak product. It has earned its place.

Module breadth. Odoo ships 30+ modules covering accounting, CRM, inventory, manufacturing, ecommerce, HR, marketing, project management, and field service. The breadth is real and most of it is usable.

Active community. Odoo has a 17 year head start. Thousands of community apps, an active forum, and a network of partners on every continent.

Decent UI. The Odoo web client is genuinely well designed for an ERP. Most ERP user interfaces look like they were drawn in 2003. Odoo’s does not.

Manufacturing depth at the Enterprise tier. Odoo MRP, work orders, quality, PLM, and shop floor are mature. If you need a real manufacturing ERP and you are willing to pay Enterprise, Odoo gets you there.

Localization library. Odoo has accounting localizations for 60+ countries, including statutory tax forms and chart of accounts templates. ERPClaw is US-first in v1.

These are real strengths. The question is whether they justify the structural tradeoffs that come with the Community vs Enterprise split.

## The Community edition tax

Here is the thing the Odoo marketing pages are quiet about. Odoo Community is the free open-source edition. Odoo Enterprise is the paid edition. The features that move from “nice demo” to “actually production-ready” mostly live in Enterprise.

Things that are Enterprise-only or significantly weaker on Community:

- Studio (the no-code customizer)

- The mobile app

- VoIP integration

- E-signature

- Marketing automation

- Document management

- Quality, PLM, MRP work orders (manufacturing depth)

- Shop floor view

- Field service

- Several accounting features including bank statement auto-reconciliation

- Helpdesk

- Subscriptions module

- Sign module

- Many country-specific localizations beyond the basic chart of accounts

You can technically run Odoo Community for free on your own server. What you get is the bones. The product most people demo at sales events, the one with the smooth onboarding, the polished mobile app, and the decent accounting workflow, is Enterprise. And Enterprise runs $31.10 to $61.00 per user per month, billed annually, with the first-year discount on Custom hiding the year-two price.

A 10 person company on Standard pays $3,732 a year. A 10 person company on Custom pays $7,320 a year (after the year-one $49.00/user discount expires). A 50 person company on Custom is at $36,600 a year. That is before partner fees, which on Odoo Enterprise routinely run $20,000 to $80,000 for a real implementation.

ERPClaw has no Community vs Enterprise split. There is one edition. It is open source. Every feature ships in the free download. Every module installs from one command. There is no upsell path because there is no upsell to sell.

## What ERPClaw replaces, in one shared database

ERPClaw covers the full back office in one install, with no per-user fee, with no Enterprise tier behind it:

- Accounting: double-entry GL, a US chart of accounts out of the box, AR/AP aging, trial balance, P&L, balance sheet, cash flow, multi-company, period close. Submitted GL entries are immutable. Every cross-table write is one transaction. Every posting passes a full validation pipeline before it touches the books.

- Inventory: items, warehouses, batches, serial numbers, reorder levels, FIFO and weighted average costing, transfer orders, stock revaluation.

- Manufacturing: bill of materials, routing, work orders, production planning, MRP.

- HR and Payroll: employees, time off, expense claims, salary structures, FICA, federal and 50 state withholding, W-2 generation. In the core, not an add-on.

- CRM: leads, opportunities, pipelines, contacts.

- Projects and Billing: project P&L, time entries, recurring invoices, usage-based billing, ASC 606 revenue recognition.

- Stripe deep integration: full event coverage, three-layer payout reconciliation, ASC 606 engine, Connect platform fees.

- Shopify deep integration: full order and payout coverage, OAuth pairing, GDPR webhooks.

- Industry verticals: retail, restaurant, healthcare, legal, nonprofit, education, real estate, agriculture, automotive, food, hospitality, construction, fleet, logistics, and more.

That entire list is one install, one database, zero per-user fees. Total this as separate Odoo Enterprise apps for a 50 person company and you are well past $30,000 a year before implementation. ERPClaw is $0 forever (open source), self-hosted on your own infrastructure.

## The AI-native difference

This is where the structural point lives, and it is the part of the comparison Odoo cannot reach.

Odoo was started in 2005 as TinyERP, written in Python on top of PostgreSQL with a custom ORM. The architecture is solid for what it is. It is also a 20 year old codebase that predates the AI era by a decade and a half. Odoo has added AI features in recent releases (chatbot helpers, document scanning, lead enrichment), and they are reasonable bolt-ons. They are also bolt-ons. The core data model, the workflow engine, the reporting layer, and the developer tooling were all designed when “AI” meant a CRM rule engine.

ERPClaw was built in 2026 with AI-native architecture from line one. What that means in practice:

- AI-native data model: every table has a stable schema with clear semantics that an AI agent can reason about without needing a 200 page training corpus first. The action surface is the API. There is no “expert mode” needed for an AI to operate the system.

- Spec-first development: every action ships with a SKILL.md describing what it does, what tables it writes, and what GL entries it produces. AI agents can read the spec and operate the action without prompt-tuning.

- Self-improving engine: ERPClaw OS is the meta-layer that catches breakages with constitutional tests, contract tests, smoke tests, and an invariant engine. The product gets safer with every release because the test suite refuses to let regressions ship.

- Clean integration story: an AI agent can sync 50 customers from Stripe, post a journal entry, run a payout reconciliation, and generate a P&L without needing a custom connector for each step. The actions are first-class.

Odoo can ship AI-decorated features. Odoo cannot ship AI-native architecture without rewriting the product. That is the structural moat.

Our team has spent years rolling out enterprise ERP and three years adjacent to NetSuite implementations. I will tell you the same thing I tell every “AI feature” demo: bolting AI onto a 2005 codebase is a UI layer, not an architecture. ERPClaw is the opposite.

## Database: SQLite default, PostgreSQL when you outgrow it

One detail people get wrong. Odoo runs on PostgreSQL. ERPClaw runs on either.

ERPClaw is database-agnostic via PyPika, a SQL query-builder abstraction. SQLite is the default because it ships zero-install with Python and runs the entire suite at small business volume on modest hardware (your laptop, a small server, your own cloud, your call). PostgreSQL is fully supported as an alternative backend. Same code, same actions, same modules. Switch by changing one connection string.

This is not a “we will support Postgres someday” promise. PyPika handles the SQL dialect translation today. ERPClaw v3.4 includes the PostgreSQL backend. Some early customers run on SQLite for the install simplicity and switch to PostgreSQL when they cross a threshold (multi-region writes, 100+ concurrent users, dedicated DBA).

The structural point: ERPClaw lets you start on SQLite (no server tuning) and scale to PostgreSQL (full ACID horizontal scaling) without changing code. Odoo requires PostgreSQL from day one and the operational overhead that comes with it.

## Pricing comparison, real numbers

Let me put the prices side by side for a 10 person company that wants a real ERP.

Odoo Enterprise Standard tier:

- 10 users at $31.10/user/month, billed annually

- $3,732 per year in license

- Implementation partner: typically $15,000 to $40,000 one-time

- Hosting: included in Online plan or self-hosted on your server

- Year 1 total: roughly $18,000 to $44,000

Odoo Enterprise Custom tier (the one with Studio):

- 10 users at $49.00/user/month for year one, $61.00/user/month after

- $5,880 in year one, $7,320 a year after that

- Implementation: same range

- Year 1 total: roughly $20,000 to $46,000; year 2 onwards: $7,320 in license alone

ERPClaw:

- License: $0 forever (open source)

- Implementation: install in 5 minutes via OpenClaw

- Hosting: self-hosted on your own infrastructure

- Year 1 total: $0 in software fees

The gap is not 2x or 3x. It is 50x to 100x. And ERPClaw does not put your essential features behind an Enterprise tier you cannot buy out of.

## Where Odoo still wins

I want to be honest about the cases where Odoo is the better pick.

You need 60+ country localizations today. ERPClaw ships regional coverage for the US, Canada, the EU, the UK, and India, with 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED) on the payment side. Odoo’s two decades of country localizations run far longer than that list. If you operate in 20 countries today, Odoo wins this row.

You have an existing Odoo partner mid-implementation. Switching ERP mid-rollout is a bad idea. Finish what you started.

You need the field service or PLM modules at depth. Odoo’s manufacturing depth at the Enterprise tier (MRP work orders, shop floor view, PLM versioning) is more mature than ERPClaw v3 ships. ERPClaw covers the manufacturing primitives (BOM, routing, work orders, MRP) but not the PLM or shop floor view.

You want a polished mobile app today. Odoo Enterprise has one. ERPClaw is CLI-first with a webclaw web UI in active development. If “I need to approve POs from my phone tomorrow” is a hard requirement, Odoo wins.

These are real cases. If you are in one of them, Odoo is the right choice. If you are not, the structural advantages of ERPClaw (no per-user fee, no tier lock-in, AI-native, open source license, the full operations surface in one DB) compound year after year.

## How to switch from Odoo to ERPClaw

If you read this far and want to try ERPClaw next to your current Odoo install, here is the path. You do not have to migrate to evaluate.

Step 1: install ERPClaw on a test machine. Five minutes via OpenClaw. The full suite. No partner needed. See the install guide.

Step 2: import your chart of accounts. ERPClaw ships with a US default chart of accounts. If your Odoo COA differs, the `import-chart-of-accounts` action takes a CSV.

Step 3: import customers, vendors, items. Odoo exports each of these to CSV from the list view. ERPClaw imports from CSV via the corresponding `import-*` actions.

Step 4: import open balances. Trial balance as of cutover date, AR aging, AP aging, inventory on hand. ERPClaw posts the opening journal entries via `import-opening-balances`.

Step 5: connect Stripe and Shopify. Both integrate in under five minutes. The Stripe app is on the Stripe Marketplace. The Shopify app is installable today; start at the feature tour.

Step 6: run parallel for a month. Post the same transactions in both systems. Compare the trial balance at month end. When the numbers tie, cut over.

This is the same parallel run that any responsible ERP migration uses. The difference is ERPClaw costs $0 to install for the parallel run, so the test costs nothing.

## What this saves you over five years

Take a 25 person company. Odoo Enterprise Custom at $49.00 per user per month in year one and $61.00 per user per month thereafter is $14,700 in year one and $18,300 every year after. Five years of license is roughly $87,900. Add a $30,000 partner implementation in year 1 and roughly $5,000 a year in customization or Studio work. Five year total: about $143,000.

Same company on ERPClaw. Five year license: $0. Five year hosting: about $1,200. Five year total: roughly $1,200.

The $142,000 difference is what AI-native open source actually costs vs what Odoo Enterprise actually costs. Multiply across the next decade and it is the difference between “we built a profitable business” and “we paid Odoo a lot of money.”

## Try ERPClaw next to your current Odoo install

The cleanest way to test the Odoo alternative claim is to run them side by side. ERPClaw installs in five minutes, costs $0 to evaluate, and you can run it on the same dev machine you use for Odoo.

- Install ERPClaw (5 minutes)

- Try the demo (browser, no install)

- See the pricing page (it is short, because it is $0)

- Read why we built an AI-native ERP

- See the QuickBooks alternative comparison (related buyer journey)

- See the ERPNext comparison (the other open-source ERP people consider)

ERPClaw is open source. The code is on GitHub. The product is free. The roadmap is public. The bet is that AI-native architecture, no per-user pricing, and an open license compound into a better product than the 20 year old codebase Odoo is asking you to rent.

## FAQ

### Is ERPClaw really free or is there a paid tier I will hit?

ERPClaw is open source. There is no paid tier. There is no Enterprise edition. Every module ships in the free download. You self-host on your own infrastructure. We may eventually offer a managed cloud version (a “we host it for you” service), but the open-source software stays free forever.

### How does ERPClaw compare to Odoo Community specifically?

Odoo Community is the free tier of Odoo. It is missing Studio, the mobile app, several accounting features, helpdesk, subscriptions, manufacturing depth, and most country localizations beyond the basic chart of accounts. ERPClaw ships the full operations surface in one tier. There is no Community vs Enterprise split.

### Can ERPClaw run on PostgreSQL like Odoo does?

Yes. ERPClaw is database-agnostic via PyPika. SQLite is the default for install simplicity, but PostgreSQL is fully supported as an alternative backend. Same code, same actions, same modules. Switch by changing one connection string. This is not a roadmap promise, the PostgreSQL backend ships in v3.4.

### What about Odoo’s 17 year head start on modules?

Odoo’s breadth is real. ERPClaw covers the same primitives from books to inventory to payroll but is younger. The case for ERPClaw is not “we have more modules than Odoo.” The case is “we have AI-native architecture, no per-user fee, no Enterprise lock, and the breadth is enough for the small to mid-market company that does not need 60 country localizations on day one.”

### Does ERPClaw have a partner network for implementation?

Not yet, and we are deliberately not building one in v1. ERPClaw installs in five minutes. Most small businesses can self-install. We may build a certified partner program later for mid-market customers who want hands-on help, but the product is built so a small business can run it without one.

### Is there a migration tool from Odoo to ERPClaw?

There are CSV import paths for chart of accounts, customers, vendors, items, and opening balances. There is not yet a one-click “migrate from Odoo” tool. The CSV path takes a competent operator a day. We are tracking demand for a deeper migration tool and may build one in 2027.

### Will ERPClaw eventually charge per user the way Odoo does?

No. The whole point of GPL v3 licensing is that the software is free forever. The business model is potentially a managed cloud product (where we host and operate it for you), professional services, and partnerships. The self-hosted open-source ERPClaw stays $0 per user per month, on every release, forever.

Install ERPClaw now or run the live demo and see if the Odoo alternative claim holds up against your real workflow.

---

## Odoo Enterprise AI Pricing 2026: Plans, Features, and Cost Math
URL: https://www.erpclaw.ai/blog/odoo-enterprise-ai-pricing-2026/

# Odoo Enterprise AI Pricing 2026: Plans, Features, and Cost Math
A dated Odoo Enterprise AI pricing snapshot with plan scope, feature access, and one, three, and five year cost math for small teams.

A founder I was talking with last week made a sensible-looking choice. She wanted an open-source ERP. Odoo by Odoo S.A. has a Community edition that is genuinely free, so she installed it on a small server, walked her team through Sales and Inventory, and was about to send a thank-you note to the open-source community. Then she went looking for the AI features she had seen in the Odoo 19 demos. None of them were in her install.

That is not a bug. Odoo’s AI features are an Enterprise-only product. Community is free, but the AI surface lives behind a $31.10/user/month paywall (US Standard, as of June 2026), and the line between the two editions is sharper in 2026 than it has ever been.

Pricing basis: US Standard plan, annual billing, captured June 2026. Taxes, currency, negotiated discounts, implementation, hosting choices, and later price changes are outside this snapshot.

This article owns the dated pricing and entitlement analysis. Use the Community vs Enterprise guide for edition selection and the Odoo alternative guide for the wider fit decision.

This post walks through exactly what that $31.10 buys on the AI side, what the math looks like over one, three, and five years for typical small teams, and what your options are if you want native AI without the per-seat fee.

## The Odoo Community vs Enterprise split

Odoo ships two editions of the same core product. The split is well documented and the OEC.sh guide to AI for Odoo Community Edition lays it out in detail.

Community. Free, open source under LGPL v3. You get accounting, sales, CRM, inventory, purchase, manufacturing basics, the website builder, and the rest of the classic Odoo footprint. You do not get the native AI features. You do not get the studio app, the dashboard editor, the advanced HR modules, or several Enterprise-only finance features either, but the AI gap is what this post is about.

Enterprise. Paid, proprietary on top of the Community core. $31.10/user/month is the published 2026 list price (US Standard). Annual billing. This is where the AI sits.

The same code base powers both editions; the differentiating modules are loaded only when you have an Enterprise subscription. Community installs can technically import some Enterprise modules through community-maintained forks, but support and stability vary, and the official AI modules are not part of the forks in any production-ready form.

## What Odoo 18 Enterprise ships natively for AI

Odoo 18 was the first release where AI moved from “available as a third-party app on the Odoo marketplace” to “part of the Enterprise bundle.” Here is what is in the box as of Q1 2026. The full list lives in the Odoo 19 AI documentation, and there is a useful catalog in the Braincuber 2026 pricing piece.

- Predictive Lead Scoring. Trains on your historical CRM data to predict which open leads are most likely to close. Sits inside the Sales module.

- AI Document OCR. Uploads invoices, receipts, and purchase orders, extracts vendor, line items, totals, dates. Hooks straight into Accounts Payable.

- AI Fields. Lets you add a free-form text field anywhere in Odoo and have the AI fill it from the surrounding record. Useful for product descriptions, summary fields, follow-up notes.

- Semantic Search. Search by intent across records rather than by exact field match. “Customers in California who churned” returns a result set even if “churned” is not a field anywhere.

- AI App for sales forecasting and stock demand. Time-series forecasting for revenue and inventory replenishment.

Every one of those features is gated to Enterprise. A Community install will see the menus but the actions will be inert or absent.

## OdooBot and the ChatGPT dependency

OdooBot is the chat assistant inside Odoo’s Discuss interface (Odoo’s internal messaging app). It is technically available in both editions, but the useful AI behavior, asking questions in natural language and getting responses about your Odoo data, depends on a third-party OpenAI integration that Odoo S.A. configures for Enterprise customers and that Community users have to wire up themselves with their own OpenAI API key.

The honest reading: OdooBot in Community is a chatbot stub. OdooBot in Enterprise is a chatbot connected to ChatGPT with your data in the prompt. The dependency on OpenAI is worth flagging on its own, because it means your Odoo data is leaving Odoo’s infrastructure on every AI call. Some buyers will be fine with that; some will not.

## Odoo 19 and the roadmap

Odoo 19 was released in late 2025 and is the current version through 2026. The headline AI addition is RAG-based agents, which means agents that can ground their answers in your Odoo data using retrieval before generating. The agents are Enterprise-only.

Odoo 20, planned for the second half of 2026, is expected to expand the agent footprint to handle multi-step actions across modules. The roadmap calls those features “agentic AI” and they are also Enterprise-only based on Odoo’s published preview materials.

## What Odoo Enterprise actually costs over time

The list price is easy. The total bill depends on how many users you have and how long you stay. Here is the simple arithmetic.

5 users.

- Year 1: $31.10 x 5 x 12 = $1,866

- Year 3 cumulative: $5,598

- Year 5 cumulative: $9,330

10 users.

- Year 1: $3,732

- Year 3 cumulative: $11,196

- Year 5 cumulative: $18,660

25 users.

- Year 1: $9,330

- Year 3 cumulative: $27,990

- Year 5 cumulative: $46,650

These numbers assume the list price holds steady, no discounting, and no add-on Odoo apps purchased separately. Real bills tend to be higher once you factor in Odoo.sh hosting (if you use it), additional implementation services, and any third-party modules you license through the Odoo App Store.

Two notes that often get missed:

- Per-user, not per-seat. Every named user counts, including admin accounts and the founder’s account. You cannot save money by having ten people share five logins; Odoo’s terms of service prohibit it.

- Annual billing locks the count. If you grow from 10 to 15 users mid-year, you owe the prorated difference on the next renewal.

A 5-person business looking at Odoo Enterprise for the AI features is making a $1,866/year decision in Year 1 and a $9,330 cumulative decision over 5 years. That is not enormous money. It is also not zero, and it is specifically the cost of unlocking AI features that the open-source license does not cover.

## Open-source alternatives that do not tier-gate AI

The open-source ERP space has split into three camps on the AI question.

Camp 1, no AI. Akaunting and Manager.io ship as open-source accounting tools with no native AI. Cheap, focused, no per-user fees, but you bring your own AI story.

Camp 2, AI as a third-party app marketplace. ERPNext by Frappe Technologies sits here. The core does not ship AI; the Frappe app marketplace has a handful of third-party AI apps you can install and license separately. Coverage varies by app, integration is uneven, and the licenses do not all match the core GPL. The full breakdown is in ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared.

Camp 3, AI-native. ERPClaw sits here. Honest disclosure, I build it. AI is the primary interface from line one, not a sidebar bolted on later. There is no Community/Enterprise split because there is no Enterprise tier; the whole product is GPL v3, free forever, self-hosted on your own infrastructure. The same install includes the action layer that lets an AI agent operate the system natively, the cryptographically signed module manifest, and the audit-grade accounting (decimals not floats, immutable GL postings, full validation on every entry).

## How to think about the trade in 2026

If you are picking between Odoo Community and Odoo Enterprise specifically for the AI:

- Stay on Community if you can live without native AI for now, or if your AI needs are narrow enough to satisfy with a third-party plugin from the Odoo App Store.

- Pay for Enterprise if you want Predictive Lead Scoring, AI Document OCR, AI Fields, Semantic Search, and the upcoming RAG agents to all “just work” inside the same UI your team already uses. The $31.10/user/month is the price of that integration.

- Look outside the Odoo ecosystem if the per-user fee bothers you and you want AI-native without a tier wall. That is where ERPClaw and a few other newer products live.

The full head-to-head between ERPClaw and Odoo, including the modules-and-coverage breakdown that this post does not duplicate, is at /compare/odoo/.

Two related reads on competitor AI stories in 2026:

- SAP Joule and SAP Business One: What’s Embedded vs What Isn’t

- ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared

## The summary

Odoo Community is genuinely free and a strong open-source ERP. The AI features that show up in 2026 Odoo demos are Odoo Enterprise features, gated at $31.10/user/month. That is roughly $1,900/year for a 5-person team and roughly $9,300/year for a 25-person team. If that fits your budget and you like the Odoo UX, Enterprise is a defensible choice. If it does not, the open-source alternatives that put AI in the free tier are a small but growing list, and they are worth a serious look before you commit to a per-seat fee.

Compare ERPClaw and Odoo side by side.

Tagsodooodoo-enterpriseai-erppricingopen-source-erp
Share:XLinkedInHacker News

---

## One Writer Per Table: Write Ownership in an ERP
URL: https://www.erpclaw.ai/blog/one-person-holds-the-pen-write-ownership-in-an-erp/

# One Writer Per Table: Write Ownership in an ERP
Read-many, write-one as an architecture rule: any module may read any table, but only the owning module may write to it. Why that makes bugs findable.

Short answer. Any module in the system may read any table. Each table has exactly one module allowed to write to it. That single constraint decides how long it takes to find a bug and how far a bad one can travel. Disclosure: we build ERPClaw and this is one of its load-bearing rules, so weigh the framing accordingly.

## The Google Doc problem

Picture a document eleven people can edit. Something in it is wrong. To find out how it got that way you open the version history and start guessing: who changed this, when, and were they reacting to somebody else’s change.

Most enterprise systems are that document. Several modules can write to the same table because at some point each of them had a good reason, and the reasons accumulated. Then a sales invoice comes out wrong and the investigation begins by asking which of eleven code paths touched it.

The cost is not the bug. The cost is that finding the bug requires understanding the whole system rather than one part of it, and that cost is paid on every bug, forever.

## The rule

Two halves, and both matter.

Any module may read any table. Reads are free and unrestricted. A module that needs to know a customer’s credit limit or an item’s on-hand quantity simply looks. No permission negotiation, no data duplication, no synchronisation job that runs at 3am and sometimes fails.

Exactly one module may write to each table. The owning module and nothing else. Where another module needs a change made, it asks the owner rather than reaching in.

Read-many keeps the system integrated. Write-one keeps it debuggable. Systems that get this wrong usually get it wrong in one direction: locking down reads, which produces data duplication and reconciliation work, or opening up writes, which produces the Google Doc.

## What it buys, concretely

One place to look. A sales invoice is wrong, so you open the module that owns sales invoices. Not a search across the codebase, not a trace through eleven candidates. One file. The time between “this is wrong” and “here is the line” collapses, and that gap is where most debugging time actually goes.

Contained blast radius. A defect in the healthcare module cannot corrupt the sales ledger, because it never had write access to it. Not through policy or code review or a test that someone remembered to write, but because the permission does not exist. This is the property that makes generated modules safe to add: a new vertical arrives with its own tables and no ability to damage anything that was already working.

Reviewable change. When a table has one writer, the diff that changes how it is written is in one place, and the reviewer knows what they are looking at. When it has six, the reviewer has to reason about interactions they cannot see.

## What it costs, because it does cost something

Three honest costs.

Cross-module operations need a path. When a workflow spans several modules, one of them owns each write, and the sequence has to be expressed. That is more design work up front than letting everything write everywhere.

The ownership map has to be maintained. A table whose owner is ambiguous is worse than no rule at all, because people will assume and the assumption will be wrong. Ownership has to be explicit and it has to be enforced, not documented and hoped for.

Some genuinely shared entities are awkward. Occasionally two modules both have a legitimate claim to write something, and resolving it means either splitting the table or accepting that one owns it and the other asks. Neither is free. We wrote about a narrow exception in the same spirit: a foundation table may carry a nullable reference to an entity another module owns, and only the owner may populate it.

## Why an AI-run system needs it more, not less

The rule is good practice in any system. It becomes structural when an AI agent is doing the writing.

A human developer who reaches into another module’s tables is doing something they know is a shortcut. They remember it, and someone catches it in review. An agent generating a module from a specification has no such instinct; it will write wherever the schema permits. So the permission model has to be the constraint, because the discipline cannot be.

That is the same reasoning behind checking invariants after every ledger operation rather than trusting the code that just ran. In both cases the system assumes the writer might be wrong and arranges things so that being wrong is survivable. We wrote about the invariant half of that argument in why an AI cannot be your accounting system on its own, and about what the ownership rule makes possible in adding an industry is a recipe, not retraining.

## The verdict

Read-many, write-one is not an elegant idea. It is a boring constraint that makes a specific bad afternoon shorter: the one where something is wrong in the books and nobody can say which part of the system did it. In a system where the writer is an agent rather than a person, boring constraints are the ones worth having.

This is one of the explanations we developed for the IEEE International Conference on Information Reuse and Integration in July 2026, where it drew more questions than we expected.

Tagsarchitectureownershiperpai-nativedebugging
Share:XLinkedInHacker News

---

## Open Source AI Accounting: The 2026 Honest Roundup
URL: https://www.erpclaw.ai/blog/open-source-ai-accounting-2026-roundup/

# Open Source AI Accounting: The 2026 Honest Roundup
2026 roundup of open source AI accounting tools. ERPClaw, ERPNext, Akaunting, Manager.io, Odoo. Which is AI-native, which is bolt-on, honest gaps.

Search “AI accounting software” in 2026 and the top of the page is paid SaaS with an AI label slapped on a product designed before transformers existed. Search “open source AI accounting” and the result set is a lot smaller, and most of what is there is not actually AI-native, just open source with a chat box plug-in. This post is the honest roundup of what open source AI accounting actually looks like in 2026, who the five candidates are, and where each of them falls short.

If you want the broader category framing first, the pillar pages cover AI accounting, open-source AI accounting, and the architecture argument at AI-native ERP. This post sits underneath those and goes deeper on the tool-by-tool comparison.

## What “open source AI accounting” actually requires

Most tools that show up under this search fail one of three tests. To count as open source AI accounting in 2026, a tool needs all three:

Test 1: the source is observable. GPL v3, MIT, AGPL, LGPL, or another OSI-approved license. You can read the code, fork it, audit it, run it on your own hardware. “Free to download” does not count. “Source-available” with a non-commercial clause does not count. The architecture has to be visible to anyone, not just paying customers.

Test 2: self-host is a first-class option. You can run it on your laptop, your VPS, your air-gapped server. There is no required cloud dependency, no phone-home, no license server that needs an internet connection. Your books live on hardware you control.

Test 3: AI-native, not AI-decorated. The AI is the primary user of the data model and the action layer, not a chat sidebar bolted on an existing UI. The agent posts journal entries directly, with invariants enforced at write time. A vendor that added a plug-in marketplace where one of the plug-ins is a chat box does not pass this test. The architecture has to be designed with the AI as the writer.

Five tools came up across the open source AI accounting category in our 2026 sweep. Here is how each one scores against the three tests.

## ERPClaw

The one we build. GPL v3, full source on github.com/avansaber/erpclaw, self-hosted by design (SQLite default, PostgreSQL fully supported via PyPika), AI-native from line one. The action layer is the AI’s API. Type a sentence, the agent invokes the right actions, the journal entries post in one transaction with a full audit row. Every posting runs through GL invariants (debits equal credits, accounts exist, period is open, no future-dated entries) before it touches the books.

Scope is full ERP plus industry verticals across retail, restaurant, healthcare, legal, nonprofit, and more. Stripe is live on the Stripe Marketplace; Shopify ships at v1.1.3 with OAuth Token Exchange via App Bridge. ASC 606 revenue recognition is built into the Stripe integration. Multi-currency support (USD, EUR, GBP, CAD, INR, SGD, AED), invoice currency equals payment currency so there is no FX guessing in the books.

Pricing is $0 forever. Not a freemium ladder, not a community-edition-with-paid-pro split. Every action across every module ships open under GPL v3.

Tests: Source observable (GPL v3), self-host (SQLite or PostgreSQL on your hardware), AI-native (action layer is the AI’s API). Passes all three.

Honest gaps. The webclaw web dashboard is developer-only today; the primary interface is an AI agent in your terminal. AICPA preferred-provider status is not something an open source project applies for. The third-party ecosystem is younger than NetSuite’s SuiteApp marketplace or QuickBooks’ add-on store. The deeper head-to-heads against mid-market commercial peers are at ERPClaw vs NetSuite, vs Sage Intacct, vs Rillet, and vs QuickBooks.

## ERPNext

ERPNext by Frappe is the open source ERP default. Frappe Cloud as the managed hosting layer, ERPNext as the application. License is GPL v3, source on Frappe’s GitHub, self-host is supported and well-documented. So far so good on tests one and two.

The wedge is test three. ERPNext does not ship an AI-native architecture; the AI capability comes from plug-ins (changAI, NextAI, Composio, Ollama-based custom integrations). The underlying ERPNext data model and workflow engine were designed without the AI as a first-class user. A chat box sits next to a forms-and-approval product; the AI suggests, a human still drives every meaningful write. That is the textbook definition of AI-decorated.

Tests: Source observable (GPL v3), self-host (yes), AI-native (no, plug-in based). Passes two of three.

This is a real product with real users; it just isn’t AI-native. If your team is already on ERPNext and you want to add an AI layer, the plug-in route works. If you are evaluating fresh and the AI-native architecture matters to you, the side-by-side detail is at ERPClaw vs ERPNext.

## Odoo Community

Odoo Community is the LGPLv3 open source edition of Odoo; Odoo Enterprise is the proprietary edition with the AI features. The split matters here. Community is self-hostable and you can read the source; Enterprise gates the meaningful AI capability behind a per-user-per-month subscription that starts at $31.10 per user per month (US Standard, as of June 2026) plus app fees, detailed in our Odoo Enterprise AI pricing breakdown.

So the question is which Odoo you mean. Odoo Community: passes test one (LGPLv3) and test two (self-host), fails test three (the AI you want lives in Enterprise). Odoo Enterprise: fails test one (proprietary), often-cloud-only by default (fails or partially fails test two), and the AI capability inside Enterprise is still mostly chat-and-suggest, not agent-posts-directly (fails test three on the AI-native criterion).

Net: Odoo doesn’t cleanly land in the open source AI accounting category. The right comparison is ERPClaw vs Odoo.

Tests (Community): Source observable (LGPLv3), self-host (yes), AI-native (no, AI gated behind Enterprise). Passes two of three with a footnote.

## Akaunting

Akaunting is open source bookkeeping under GPL v3 with a paid SaaS layer on top. Source on GitHub, self-host supported, hosted Akaunting Cloud as the commercial offering. Good open source posture; passes tests one and two cleanly.

The wedge is test three. Akaunting does not market an AI agent as a core feature today. The product is a competently-built double-entry bookkeeping tool with apps for invoicing, expense tracking, banking, and similar. AI is not the architectural choice; bookkeeping is. That is fine. It is just not what this roundup is about.

Tests: Source observable (GPL v3), self-host (yes), AI-native (no, not the design center). Passes two of three.

If your team wants open source bookkeeping without the AI architecture, Akaunting is a sensible pick. If the AI-native architecture matters, you are in the wrong roundup. The AI bookkeeping pillar goes deeper on what AI-native means in the bookkeeping cut specifically.

## Manager.io

Manager.io is the awkward one. The desktop version is free, but the source code is not open; the Cloud and Server editions are paid. Calling Manager.io “open source” is the common shorthand, but the technically-accurate label is “freeware desktop with closed source.” It runs on your machine if you pick the desktop edition, so it scores well on the spirit of self-host even though test one (source observable) does not pass.

On test three, Manager.io is bookkeeping software; the AI angle is not the design center.

Tests: Source observable (no, freeware not open source), self-host (yes via desktop), AI-native (no). Passes one of three with a footnote.

If your evaluation is “free desktop bookkeeping that runs on my laptop,” Manager.io is a real choice. If your evaluation is “open source AI accounting,” Manager.io does not actually clear the bar.

## The scoreboard

| Tool
| Source observable
| Self-host
| AI-native

| ERPClaw
| Yes (GPL v3)
| Yes
| Yes

| ERPNext
| Yes (GPL v3)
| Yes
| No (plug-in based)

| Odoo Community
| Yes (LGPLv3)
| Yes
| No (AI gated in Enterprise)

| Akaunting
| Yes (GPL v3)
| Yes
| No (not the design center)

| Manager.io
| No (freeware)
| Yes (desktop)
| No

ERPClaw is the only tool in the 2026 open source AI accounting category, at least that we have found. We hold this claim hedged because the open source landscape moves fast; if you know a tool we missed, file an issue on the ERPClaw repo and we will update the roundup.

## What this means if you are evaluating

Three practical takeaways.

If you want AI-native open source today, ERPClaw is the answer this roundup converges on. The architecture choice (AI as the action layer’s primary user) is not retrofittable onto a 1990s general ledger or a forms-and-approvals ERP. ERPNext, Odoo, Akaunting, and Manager.io would each need a foundational rewrite to claim the AI-native label honestly.

If you want open source bookkeeping without the AI architecture, Akaunting and Manager.io are real options. Pick based on whether source-observability matters to your audit posture (Akaunting wins) or whether desktop simplicity matters more (Manager.io wins).

If you want open source ERP with optional AI plug-ins, ERPNext is the most mature option. The plug-in route works; you just trade the AI-native architecture for an AI-decorated one.

The architecture argument is at open-source AI accounting; the broader AI-native framework is at AI-native ERP; the specific bookkeeping and inventory cuts are at AI bookkeeping and AI inventory. If you are coming from QuickBooks, the switching guide is at AI for QuickBooks.

## Frequently asked questions

Is there a free AI accounting tool besides ERPClaw?

For “free” defined as “open source plus AI-native,” our 2026 sweep returned ERPClaw and no other clean answers. For “free” defined as “freemium tier with AI features,” vendors like Puzzle and Wave offer free tiers; both are closed source and not AI-native. If you know an open source AI-native option we missed, let us know via the ERPClaw repo.

Why doesn’t ERPNext count as AI-native if it has AI plug-ins?

Because the AI is added through a plug-in marketplace on top of a data model and workflow engine designed without the agent as the primary user. The plug-ins suggest categorizations and answer questions about the books; they do not autonomously post journal entries with invariants enforced at write time. That is the AI-decorated pattern. The pillar at open-source AI accounting breaks down the 5-trait test we use to draw the line.

What about Odoo Enterprise’s AI features?

Odoo Enterprise is proprietary, not open source. It also gates the meaningful AI capabilities behind a per-user-per-month subscription. Even setting the open source question aside, the AI architecture in Enterprise is still chat-and-suggest rather than agent-posts-directly. The honest comparison is at ERPClaw vs Odoo.

Can ERPClaw replace QuickBooks?

For most small to mid-market use cases, yes. ERPClaw covers accounting (immutable double-entry GL), invoicing, payments, payroll, inventory, US tax forms, plus a library of industry verticals. The migration playbook is at /migrate/from-quickbooks/ and the head-to-head is at ERPClaw vs QuickBooks. What we do not have today is a polished web dashboard for non-technical users; the primary interface is an AI agent in a terminal.

How long does it take to install ERPClaw?

5 minutes for the install (4 commands: install ClawHub, install ERPClaw, start the chat agent, describe your business). 15 to 30 minutes more to import your chart of accounts and opening balances if you are migrating from another system. By the end of an hour, the AI agent is posting journal entries against your live data. The full install path is at /docs/core/install/.

Is open source AI accounting safe for production finance data?

Self-hosted open source means your books live on your hardware. Nothing leaves your infrastructure unless you explicitly send it. The AI agent runs locally or in your cloud. By contrast, every SaaS AI accounting service stores your data on vendor servers; you trade local control for vendor convenience. CPAs increasingly prefer the open source posture for the same reason they accept Linux on production servers: the controls are observable.

Will my auditor or audit committee accept books from an open source AI tool?

For most small to mid-market audits, yes. ERPClaw exports trial balance, P&L, balance sheet, AR aging, AP aging, GL detail, and the core financial report set as CSV or PDF in standard formats. The chain-hashed audit trail and immutable GL produce books that pass external review cleanly. Where the answer is “not yet” is when your audit committee specifically requires an AICPA-endorsed SaaS vendor with a contracted SLA; in that case Sage Intacct and similar peers fit that procurement mandate directly. The detailed comparison with mid-market commercial peers is at ERPClaw vs Sage Intacct.

Why does the AI architecture matter if the AI works fine as a plug-in?

For light tasks like answering questions about your books or suggesting categorizations, plug-in AI works. For the work that compounds (the agent autonomously posting journal entries, reconciling Stripe payouts end to end, running the close calendar against a real ledger), the architecture matters because invariants need to be enforced at write time, not after a human approves. Bolt-on AI suggests; AI-native AI posts. If your team will adopt chat-first and you want the AI to actually do the work, the architecture is the decision. If your team will keep humans in the loop on every meaningful write, the architecture matters less and the plug-in route is fine.

## Where to go next

Free forever, install in 5 minutes, every module included. Talk to a co-founder at /demo/, browse the source at github.com/avansaber/erpclaw, or read the pillar at AI accounting for the broader category framing.

Tagsopen-sourceai-accountingcomparisonroundupself-hosted
Share:XLinkedInHacker News

---

## Open Source vs SaaS: The ERP Question Most Buyers Get Wrong
URL: https://www.erpclaw.ai/blog/open-source-vs-saas-the-erp-question/

# Open Source vs SaaS: The ERP Question Most Buyers Get Wrong
SaaS ERP vendors charge per user forever and own your data. Open source gives you the code and runs on your machine. The honest tradeoff, either way.

For example, a founder I know runs a $4M revenue ecommerce business. She pays $235 a month for QuickBooks Online Advanced, $89 a month for Synder, $79 a month for Shopify Plus apps that handle inventory, $40 a month for a CRM, and $120 a month for a payroll service. About $7,000 a year in software subscriptions for what is, structurally, an ERP. None of the data is hers. If QuickBooks raises prices 30% next year, she pays. If Synder gets acquired and the new owner kills the integration she relies on, she scrambles. If she ever wants to leave, she gets a CSV export and spends a month re-entering everything into the next tool.

Across the table from her is a different founder running the same kind of business. He installs an open-source ERP on his own infrastructure. The license is $0 and stays $0 forever. The data is in a SQLite or PostgreSQL database on his machine. If the project goes dormant, he still has the code and can fork it. If he wants to switch tools, he runs a SQL export and has every transaction since the company started, in his own format, on his own hardware.

These two founders are running the same business. One is paying about $7,000 a year and renting their books. The other is paying $0 in software fees and owns them outright.

This is the open source vs SaaS ERP question. Most buyers default to SaaS without thinking about it. This post is the honest tradeoff so you can pick on purpose.

## The structural difference, in one paragraph

SaaS ERP is a multi-tenant cloud product. You log in, the vendor hosts your data, you pay per user per month forever. Examples: NetSuite, QuickBooks Online, Xero, FreshBooks, Sage Intacct, Rillet. The vendor controls the software, the data, the upgrade path, and the price.

Open-source ERP is software you install on your own infrastructure. The source code is published under a license that lets you read, modify, and run it. The data lives in your database, which you control. Examples: Odoo Community, ERPNext, ERPClaw. The vendor (if there is one) sells optional services or hosting, but the core software is free and you own the install.

These are not adjacent product categories. They are structurally different relationships between you and your accounting system.

## What SaaS vendors are honestly good at

I want to be careful not to write a one-sided post. SaaS ERP earned its dominance for real reasons.

Zero install. Sign up, fill in your company name, start invoicing in 15 minutes. No server, no install, no IT person needed. For a solo founder who wants to send their first invoice today, SaaS is hard to beat.

Automatic upgrades. The vendor patches the software, fixes bugs, ships new features, all without you doing anything. You never have to think about “what version am I running.”

Real support. When something breaks, you call a phone number and a human helps you. Most open-source projects do not have this.

Compliance handled for you. SOC 2, GDPR, payroll filings, tax forms. The vendor (in theory) keeps up with regulations. You sign the contract and stop thinking about it.

Mobile apps. Most SaaS ERP vendors have polished iOS and Android apps. Most open-source ERPs do not, or have weak ones.

These are real benefits. If your business is small enough that the per-user fee does not hurt and stable enough that vendor lock-in is not a concern, SaaS is fine. The honest comparison only gets interesting when the math changes.

## Where SaaS pricing breaks down

The SaaS pricing model is per user per month. It works at low user counts and breaks down quickly as you grow.

The 50-person company example. Take a small but growing business. 50 employees who all need access to the ERP. NetSuite at around $99 per user per month (contact Oracle for current pricing) is around $59,400 per year for 50 users. QuickBooks Online Advanced at $235 per month plus per-user seats stops scaling and you migrate up. Xero at $80 per user per month for the top tier is $48,000 per year.

Same company on open-source ERP: $0 in license, regardless of user count. A modest hosting fee. Under $300 a year, total.

The price increase mechanism. SaaS contracts include “we may adjust pricing annually” clauses. They are not theoretical. NetSuite’s price increases on renewal have been documented at 5% to 20% per year. QuickBooks reportedly raised the Online Advanced price from $200 to $235 over multiple years (see quickbooks.intuit.com/pricing for current rates). Xero reportedly raised prices in 2023 by 7% to 14% across plans (see xero.com/pricing for current rates). Once your data is in a SaaS tool and the migration cost is six figures, the vendor has pricing power and exercises it.

The seat tax penalty for growth. When you hire your 51st employee, your ERP bill goes up by $99. When you hire the 100th, the bill goes up by another $99. Open-source ERP does not care how many users you have. The license is the license.

The math is not “SaaS is bad.” It is “SaaS pricing is structurally optimized for the vendor, not you, and the gap widens every year you stay.”

## What “you don’t own your data” actually means

Every SaaS vendor will tell you “you own your data” in marketing copy. The reality is more nuanced.

Export formats. Most SaaS vendors offer CSV export of customers, vendors, items, transactions. What they do not offer is a clean export of the GL with referential integrity, the chart of accounts mapping, the link between an invoice and the GL entries it produced, the audit trail of who changed what when. The export is the surface; the structural model is locked in.

Migration friction. I have personally migrated companies off QuickBooks, off Xero, off Sage. Each one took two to four months because the export was incomplete and the importing side did not match. The fact that the data was technically “yours” did not make it portable.

Vendor goes away or changes terms. When a SaaS vendor gets acquired, sunsets a feature, or doubles the price, you have limited options. You can pay the new price, you can migrate (months of work), or you can stop using it (lose the data). Open-source ERP, even if the project goes dormant, still runs. The code does not disappear because the company did.

Audit access. When your auditor asks for the GL detail, the SaaS export is what they get. Some vendors limit the number of years available in export. Some charge for “data retrieval” if you cancel and need to retrieve historical data. Read your contract.

The honest framing is: with SaaS, the data is yours in the sense that you can view it and export it on the vendor’s terms. With open-source self-hosted, the data is yours in the sense that the database file is on your hard drive and you can run any query you want against it.

## Where open-source ERP is honestly weak

I am not pretending open source is universally better. There are real weaknesses, and writing this post without naming them would be dishonest.

Install friction. “Run a server” is a higher bar than “sign up on a website.” Most small business owners cannot run a server themselves. The fix is either hire someone for an hour or pick an open-source product (like ERPClaw) where the install is genuinely a single command.

No phone support by default. When something breaks at 11pm on a Sunday, no one is going to pick up. You either fix it yourself, post on a forum, or pay a partner. This is a real gap.

Variable depth across modules. Open-source projects do some things well and some things badly. ERPNext has great inventory and weak payroll. Odoo Community has decent accounting and locks half the useful features behind Enterprise. ERPClaw is younger; multi-currency is USD-only in v1. You have to evaluate against your actual needs.

No dedicated security team. SaaS vendors have security teams that find and fix vulnerabilities. Open-source projects rely on community reporting and maintainer responsiveness. The track record varies.

Mobile apps. Most open-source ERPs have weak or no mobile apps. If “approve a PO from my phone” is a daily workflow, this matters.

The honest pick depends on which weaknesses are dealbreakers for you and which are tolerable.

## The AI-native angle

There is a third axis that has only become relevant in the last 18 months: how the ERP was architected with respect to AI agents and automation. We cover this in depth in our pillar on AI-native ERP.

Almost every SaaS ERP on the market today was built between 1999 and 2015. The data models, the workflow engines, the integration patterns, all predate the AI era. They have added AI features in recent releases, but those are bolt-ons. The core architecture cannot be retrofitted to be AI-native without rewriting the product.

ERPClaw was built in 2026 with AI-native architecture from line one. Every action is documented in a SKILL.md spec. Every table has a stable schema with clear semantics. Every integration is a first-class action that an AI agent can call without prompt-tuning. The whole system is designed for automation, not just for human users clicking buttons.

This is the structural moat that does not show up in a pricing comparison. Five years from now, the difference between AI-native architecture and AI-decorated bolt-ons will be the difference between an ERP that runs itself and an ERP that requires the same human button-clicking it did in 2010.

Open source plus AI-native is a stronger combination than open source alone. Most existing open-source ERPs (Odoo, ERPNext) are AI-decorated, not AI-native. ERPClaw is the rare combination.

## The 5 year cost comparison, real numbers

Take a 25 person company. Run the cost projection over 5 years.

Path A: NetSuite.

- License: $99/user/month × 25 users × 60 months = $148,500

- Implementation: $40,000 one-time

- Customizations and SuiteApps: ~$5,000/year × 5 = $25,000

- Annual price increases: ~5%/year, compounded

- 5 year total: ~$235,000

Path B: QuickBooks Online Advanced + bolt-ons.

- QBO Advanced: $235/month × 60 = $14,100

- Bolt-ons (CRM, inventory, payroll, A2X/Synder): ~$500/month × 60 = $30,000

- Annual price increases: ~10%/year

- Migration cost when QBO can no longer hold the company: $20,000+ (eventually you outgrow it)

- 5 year total: ~$60,000 to $80,000

Path C: Open-source ERP (ERPClaw).

- License: $0 forever

- 5 year total: $0 in software cost. Your only cost is the infrastructure you would have anyway.

The 5 year gap between SaaS and open source on a 25 person company is roughly $60,000 to $235,000 depending on which SaaS path. That is real money. For a bootstrapped business, it is the difference between hiring a person and not.

## When SaaS is the right pick

I am not saying every business should switch to open source. There are cases where SaaS is honestly the better pick.

Solo founder, first 30 days. You do not have time to evaluate open source. Sign up for QuickBooks Online or Xero, send your first invoice, get back to work. Revisit in year 2.

No technical capacity at all. If you cannot run a server, do not want to learn, and cannot hire someone for one hour, the install gap is real. Vendor-run SaaS handles it, or have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

Compliance-heavy industry that needs SOC 2 or HIPAA today. Most SaaS vendors carry compliance certifications. Self-hosted open source means you are responsible for compliance. If you need a SOC 2 report next month, SaaS is faster.

Outgrowing one SaaS into another’s prime range. If you are migrating from QuickBooks to NetSuite anyway, the open-source path is a third option, but if your CFO has already picked NetSuite, fighting that decision is not worth the political capital.

These are real cases. SaaS is not always wrong. It is just often not the right pick for the reasons people assume.

## What “managed open source” looks like

There is a fourth option emerging that splits the difference: managed open source. The software is open source and you can self-host if you want, OR you pay a managed cloud version where the vendor hosts it for you. You get the lock-in protection of open source (because you can leave to self-host any time) plus the convenience of SaaS (because the vendor handles ops).

Examples: GitLab (self-hosted free, GitLab.com paid), Mattermost, Sentry, ERPClaw (self-hosted free; ERPClaw Cloud Managed hosted by us, onboarding by request through /pricing/).

This is structurally different from pure SaaS because the source code is open and the data is portable to a self-host install at any time. It is the model that matches “I want SaaS convenience without the lock-in.”

## How to actually decide

If you are in front of this decision today, here is the framework I would use.

Step 1: pick your time horizon. Are you optimizing for the next 12 months or the next 5 years? Open source loses on month 1 (install friction) and wins by month 18 (cost compounds, no lock-in). If you only care about month 1, pick SaaS.

Step 2: estimate your user count in 24 months. If you are at 5 users today and will be at 5 in two years, SaaS pricing is fine. If you are at 5 today and will be at 50 in two years, the SaaS bill grows 10x and open source stays flat.

Step 3: assess your data sensitivity. If your accounting data is something you would be uncomfortable having on a vendor’s cloud (because you compete with the vendor, because you are in a regulated industry, because you have signed NDAs that limit cloud storage), open source on your own infrastructure is the cleaner answer.

Step 4: assess your technical capacity. If you have one person on the team who can run a Linux server and read SQL, open source is genuinely accessible. If you do not, open source means hiring someone or paying a managed-open-source vendor.

Step 5: try both. ERPClaw installs in five minutes and costs $0 to evaluate. Sign up for a QuickBooks Online trial. Run the same week of transactions through both. Compare the bill, the data ownership, the GL detail, the export format. Decide.

## Try the open source path

If the cost math, data ownership, and AI-native angle resonate, ERPClaw is open source, ships all included modules in one shared database, and runs on SQLite (default) or PostgreSQL via PyPika. Free forever. AI-native architecture. No upgrade tier. No per-user fee.

- Install ERPClaw (5 minutes)

- Try the demo (browser, no install)

- See pricing (it is $0)

- See the QuickBooks alternative comparison

- See the NetSuite alternative comparison

- Read about AI-native architecture

The bet is that AI-native plus open source license plus self-hosted compounds into a better product than the per-user-per-month SaaS model can survive. The 5 year math says the gap is real money.

## FAQ

### Is open-source ERP really free or is there a hidden cost?

The license is free. The hidden cost is your time to install, configure, and operate the system. For ERPClaw specifically, install is one command (about 5 minutes), configuration is a few CSV imports for your chart of accounts and customers, and operation is mostly automated. For older open-source ERPs (Odoo, ERPNext), the configuration time is meaningfully higher.

### What happens if the open-source project goes dormant?

You still have the code and the data. You can run it indefinitely on the version you have. You can fork it and patch security issues yourself or with a contractor. You can pay a third party to maintain it. None of those options exist for a SaaS product that shuts down.

### How do I get support for an open-source ERP?

Three paths. (1) Community forums, GitHub issues, Discord channels. Free, variable response time. (2) Pay a partner or contractor for hands-on help. Cheaper than SaaS support contracts at low volume. (3) Self-support if you have technical capacity. Most ERPClaw issues are debuggable from the SQLite or PostgreSQL data and the code is open.

### Is self-hosted ERP secure?

It depends on how you run it. A self-hosted ERP on a properly configured Linux server with TLS, firewall, fail2ban, and regular patching is at least as secure as most SaaS deployments and arguably more so because you are not a multi-tenant target. A self-hosted ERP on a misconfigured server is less secure. Most small businesses follow a checklist (UFW, SSH key auth, fail2ban, regular updates) and it is sufficient.

### Can ERPClaw replace QuickBooks for a typical small business?

Yes for the businesses where QuickBooks is the right shape. ERPClaw covers AR, AP, GL, payroll, inventory, and the Stripe and Shopify integrations that most QBO bolt-ons handle. The gap vs QBO is around statutory compliance for non-US jurisdictions (ERPClaw v1 is US-only) and around the polished mobile app. See the QuickBooks alternative comparison for the detailed map.

### Will I lose my data if I cancel an open-source install?

You cannot really “cancel” an open-source install. The data is in a SQLite or PostgreSQL database on your machine. If you stop running the software, the data file is still there. You can pick it up six months later and resume.

### Is the AI-native angle just marketing or does it actually matter?

It matters at the architecture level, which is the part marketing usually skips over. AI-native means the product was designed for an AI agent to operate it (every action is a documented spec, every table is queryable, every workflow is a first-class API). AI-decorated means the product has chatbot widgets bolted onto a 2005 codebase. The first will keep getting faster every quarter as AI agents improve. The second will hit a ceiling.

Install ERPClaw and see if the open-source path holds up against your real workflow.

---

## The QuickBooks Alternative That's Actually a Full ERP
URL: https://www.erpclaw.ai/blog/quickbooks-alternative-open-source-erp/

# The QuickBooks Alternative That's Actually a Full ERP
A QuickBooks alternative that is a full ERP: accounting, inventory, and payroll in one free, open-source, self-hosted system.

QuickBooks is not bad software. Let me start there, because most “QuickBooks alternative” posts open with a list of grievances and that is not the honest take. Intuit shipped the de facto SMB accounting standard in the US. Every CPA knows it, the mobile app works, the bank feeds work, the 1099 export works. If you are a freelancer with one client and a checking account, QuickBooks Online Simple Start at $38 a month is genuinely fine.

This post is not for that person.

This post is for the person who started on QuickBooks two years ago, fits comfortably in QBO Plus, and is now doing things QuickBooks was never built to do. You know who you are because you have a set of habits that did not exist a year ago. You export a CSV every Monday to figure out what stock you actually have. You added Synder so your Stripe payouts reconcile. You pay a bookkeeper extra to build a “sales by SKU by location” report because the native one does not exist. You opened a second QBO subscription for your second LLC because Intuit charges per company file. You are paying $200 to $400 a month across QuickBooks plus three add-ons and you still spreadsheet the gaps.

That is the moment you outgrew QuickBooks. Most people do not notice it because the cost crept up one line item at a time.

ERPClaw is what comes next. It is free, open source (open source), self-hosted, AI-native, and it covers accounting plus inventory plus manufacturing plus HR plus payroll plus CRM plus projects in a single system with one database. The claim is not “we do accounting better than QuickBooks.” The claim is “we do accounting and the eight other things you bolted on, in one place, for zero per-user fees, forever, operated through your AI assistant rather than a 1990s forms UI.”

QuickBooks is excellent at what it does, but it is AI-decorated. Intuit Assist is a chat helper bolted onto a 1983 general ledger product, sold for an extra fee on top of the existing subscription. ERPClaw is the only AI-native option in this category, which means the architecture, the spec-first build, and the pricing all reflect a cost base that QuickBooks cannot match without rewriting from scratch. The full argument is in AI-decorated vs AI-native software.

## What QuickBooks does well

I want to be precise about this because I am not interested in trashing a product that millions of businesses run on.

QBO is built for the small business accounting workflow. The interface assumes you are not an accountant and nudges you to the right entries without ever showing you a journal. Most ERPs (NetSuite, Sage Intacct, even Xero) assume the user is comfortable with accounting fundamentals. QBO assumes nothing. For a non-accountant founder, that lower floor matters.

The CPA network effect is enormous. When you tell your accountant “I’m on QBO” they say “great, send me the invite link.” When you say “I’m on Odoo” they say “hmm, can you export to QuickBooks.” That friction is real, and it comes from being the standard for two decades.

Bank feeds and 1099s just work in the US. QBO connects to virtually every American bank and the 1099-NEC/MISC e-file pipeline is mature.

The case for ERPClaw is not that QuickBooks is broken. It is that the QuickBooks footprint is smaller than your business is if your business has any of: physical inventory, more than 10 employees, manufacturing, multiple legal entities, a SaaS revenue model, project-based billing, or a CRM that needs to talk to invoicing.

## The QuickBooks tax

Here is the part nobody on Intuit’s marketing site spells out for you. QuickBooks is a recurring monthly bill that scales with three things: users, features, and entities.

The current US pricing on quickbooks.intuit.com/pricing (as of June 2026) for QuickBooks Online:

- Simple Start: $38/month, 1 user

- Essentials: $75/month, 3 users

- Plus: $115/month, 5 users, basic inventory

- Advanced: $275/month, 25 users, batch invoicing

That is the headline. Now look at what the real monthly bill actually adds up to for a normal small business:

- QBO Plus, $115

- QuickBooks Payroll Premium for 8 employees, $85 base plus $9 per employee = $157

- QuickBooks Time, $20 plus $10 per user = $80

- Synder for Stripe sync, $30

- A2X for Shopify, $49

- Second company file for your separate LLC, another $115

- Bill Pay add-on, $15

- A bookkeeper to handle the parts none of the above handle, $400 to $800

You are at $960 to $1,360 per month before you have done anything fancy. A year is $11,500 to $16,500. None of that is wasted exactly. You are getting work done. But every line item is a recurring rent payment to a vendor you do not own and cannot fork. Every additional user, every additional feature tier, every additional entity ratchets it up.

This is what I call the QuickBooks tax. It is not the $38 starting price. It is the seven add-ons you accumulated to make $38 work for a real business.

If your monthly software bill across QBO, Synder, Bill.com, Gusto, and a bookkeeper has crept past $400, this post is for you. That number is a tell. QBO stopped being your accounting system a while ago and started being one of five tools you stitch together.

## What ERPClaw replaces beyond accounting

This is the differentiated point and the reason I am writing this post.

A typical “best QuickBooks alternative” listicle compares other accounting tools: Xero, FreshBooks, Wave, Zoho Books, Sage. Same shape, different price.

ERPClaw is a different category. It is a full ERP that includes accounting, the same way SAP includes accounting and NetSuite includes accounting. The accounting piece is one domain among many in a single system, sharing one database, with all data queryable by any module.

You stop running a stack and start running a system. Concretely, ERPClaw includes:

- Accounting: a US chart of accounts out of the box, a real audit-grade accounting ledger your CPA will recognize, multi-company, multi-currency, AR aging (who owes you), AP aging (who you owe), trial balance, P&L, balance sheet, cash flow. See features/financial-ops.

- Inventory: items, warehouses, batches, serial numbers, reorder levels, stock entries, FIFO and weighted-average costing. Real inventory, not the toy inventory in QBO Plus.

- Manufacturing: bill of materials, production planning, work orders, routing. If you assemble anything, this exists in ERPClaw and does not exist in QuickBooks at any tier.

- HR: employees, departments, job titles, time off, attendance, expense claims.

- Payroll: salary structures, FICA, federal and state income tax withholding, W-2 generation, garnishments. Not an add-on. Same database.

- CRM: leads, opportunities, pipelines, contacts. Not Salesforce, but enough to stop paying for HubSpot Starter.

- Projects: project tracking, time entries, project-based billing.

- Billing: recurring invoices, usage-based billing, subscription management.

- Tax: tax templates, multi-rate, tax returns.

- Stripe deep integration: all core Stripe actions, three-layer payout reconciliation, ASC 606 revenue recognition, Connect platform fees. (Replaces Synder, Rillet, and the $30 to $200 a month tools that do this.) Detailed in our Stripe launch post.

- Shopify deep integration: all core Shopify actions, the same three-layer pattern, GDPR webhooks, App Store OAuth pairing. (Replaces A2X and Webgility.)

- Industry verticals: retail, restaurant, healthcare, legal, nonprofit, education, real estate, agriculture, automotive, food, hospitality, construction, fleet, logistics, and more. Install only the ones you need.

Total those bullets as separate SaaS subscriptions and you are past $2,000 per month for a 25-person company. ERPClaw is $0 forever (open source), self-hosted on your own infrastructure.

That is not just a price argument. It is structural: when accounting, inventory, payroll, and CRM all live in one database, your AR aging report knows the customer’s CRM history, your inventory valuation knows your manufacturing routings, and your project P&L knows the actual labor hours from HR. The connections between these things are not “supported,” they are inherent. There is no sync to break. No 11pm “Synder import failed” email.

## Free and open source is a structural advantage

QuickBooks cannot be free. Intuit has 18,000 employees, a sales force, and shareholders, and their pricing has to scale with you because their cost structure scales with them.

ERPClaw can be free because it does not have any of that. It is built at AvanSaber Inc by a small co-founder team (me, my co-founder Varun, and a handful of advisors), pairing with Claude Code as a coding companion. The marginal cost of one more user is zero: no cloud bill (you host it), no support tier, no per-seat sales overhead. The open source license means it stays free even if the company changes hands.

The bigger consequence: scope can grow without re-pricing. When ERPClaw adds a vertical (we just added grooming, tattoo, and storage via ERPClaw OS), you get it for $0. When QuickBooks adds a feature, it is gated to a higher tier. The pricing model is the product roadmap.

In year three, the QuickBooks bill is twice what it was in year one. The ERPClaw bill is still $0. If you want to see how that $0 stacks up against everything else that calls itself free, free AI accounting tools in 2026 runs the honest comparison.

## Side by side

The full comparison is on /compare/quickbooks. Pulling the table here for reference:

| Feature
| ERPClaw
| QuickBooks

| Price
| Free (open source license)
| $38 to $275/month

| Users
| Unlimited
| Per-user pricing

| Interface
| Natural language chat
| Web forms and menus

| Setup time
| 5 minutes
| Days to weeks

| Training
| None needed
| Hours of tutorials

| Self-hosted
| Yes (your server)
| No (cloud only)

| Open source
| Yes (open source)
| No (proprietary)

| Data ownership
| 100% yours
| Intuit’s cloud

| AI capabilities
| AI-native (every interaction)
| AI-decorated (Intuit Assist sidebar)

| Database
| SQLite or PostgreSQL via PyPika, your machine
| Intuit’s cloud only

| Inventory
| Full (batches, serials, manufacturing)
| Basic (Plus plan only)

| Manufacturing
| Included (BOM, MRP, work orders)
| Not available

| HR & Payroll
| Included
| Separate add-on

| Multi-company
| Included
| Separate subscription per company

| API access
| Full (open source)
| Limited, paid tiers

| Offline access
| Always (local database)
| Limited

Two of those rows deserve a second look.

Interface: chat vs. forms. QBO is a forms application. ERPClaw is conversational: you talk to it through Telegram, WhatsApp, the web dashboard, or any OpenClaw client. For example: “Create an invoice for Acme, $5,000 for consulting, due in 30 days.” It writes the entry, posts it to your books, and confirms. Not a chatbot wrapper around forms. It is the native interface.

Multi-company. Stings if you have ever opened a second QBO subscription for your holding LLC or your second store. ERPClaw is multi-company in the core. One install, N companies, intercompany and consolidation included.

## When you should pick QuickBooks anyway

I promised honest. Here are the cases where I would tell you to stay on QuickBooks and not even bother with ERPClaw.

You only need accounting and you have one user. If your business is “freelancer plus contractor 1099s plus a Square account” and you have no inventory, no payroll, no second entity, and no recurring SaaS revenue, QBO Simple Start is the right tool. ERPClaw would be overkill. You would never use most of the modules.

Your CPA flatly refuses to work outside QBO. Some CPAs will work with anything that produces a clean trial balance and ledger export (we do, see /docs). Some CPAs only know QuickBooks and will charge you a “switch fee” of $1,500 to learn anything else. If your CPA is the second kind and you love them, the math may not work out. Ask before you switch, not after.

You will not run your own server. ERPClaw is self-hosted by default, and self-hosted AI accounting walks through what running your books on your own hardware actually involves. If “I do not want to think about a server, ever” is a hard line, QBO handles that, or have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

You need a feature that requires an enterprise vendor. If you need NACHA-direct ACH origination as the primary payment rail, multi-state SUI tax filing across 30 states, or you process more than 10,000 transactions a day, you are in mid-market territory and you need NetSuite, Sage Intacct, or a higher tier than what either QBO or ERPClaw is targeting today.

If none of those four apply, ERPClaw is genuinely worth a weekend of your time to evaluate.

## How a migration actually works

A standalone “Migrating from QuickBooks to ERPClaw” guide is in our docs roadmap for Q3 2026. Until it ships, the short version below covers the scope.

Step 1. Export from QuickBooks. From QBO, go to Settings, then Export Data. You get a zip with chart of accounts, customers, vendors, items, invoices, bills, payments, and the ledger detail. About 10 minutes.

Step 2. Map your chart of accounts. ERPClaw ships with a US default chart of accounts. Most QBO charts are subsets of this. Our mapping script reads your QBO chart export and produces a one-to-one mapping (with a quick manual review for any custom accounts you added). 30 to 90 minutes depending on how custom your chart is.

Step 3. Pick a cutover date. Standard practice: pick the start of a month. Run QuickBooks through the end of the prior month, close it cleanly, then open ERPClaw with opening balances as of day 1 of the new month.

Step 4. Import opening balances. One command loads your trial balance into ERPClaw as opening entries. The validation engine rejects any imbalance, so you know the import is clean before you start.

Step 5. Import master data. Customers, vendors, items, employees. Each is a one-line import. About 15 minutes.

Step 6. Run parallel for one month. This is the part most migration guides skip and you should not. Run both systems for one full month, post the same transactions to both, and at month-end compare the trial balances. They should match to the cent. If they do, kill QuickBooks. If they do not, find the discrepancy first.

For a typical 25-person small business, the whole migration is one to three weekends. For a freelancer, two hours.

## FAQ

Will my CPA still be able to help me?

Yes if your CPA understands a trial balance and a ledger export, which is the same skill they already use for non-QBO clients. ERPClaw exports a clean trial balance, P&L, balance sheet, and detailed ledger in CSV or PDF. If your CPA insists on QBO write-access specifically, ERPClaw will not satisfy that today. The honest move: ask your CPA before you switch.

How does chart-of-accounts mapping work?

ERPClaw ships with a US default chart of accounts. Your existing QBO chart is almost certainly a strict subset. The mapping script matches accounts by name and type, surfaces unmatched accounts for review, and writes the mapping to your ERPClaw config. You can also bring your QBO chart over verbatim.

Does ERPClaw handle multi-currency?

Yes, in the core, not a paid tier. Transact in any ISO currency, hold balances in multiple functional currencies, run consolidation in your reporting currency. Rates can be manual or set to a daily feed.

What about payroll?

US payroll is in the core: salary structures, FICA, federal income tax withholding, state income tax for all 50 states, W-2 generation, garnishments. The piece we do not yet ship is direct e-filing of 941s and state returns. You generate the forms in ERPClaw and file via the IRS or state portal, or hand them to your existing filer.

How is inventory different from QuickBooks Plus?

QBO Plus has basic inventory: quantity on hand, reorder points, FIFO. ERPClaw has lot and batch tracking, serial numbers, multi-warehouse with transfer orders, recipes (BOMs) that consume components when you produce a finished good, production planning, weighted-average and FIFO costing per item, and stock revaluation.

Is the audit trail good enough for an audit?

Yes. The accounting ledger is locked: entries cannot be edited after they are posted, and a cancellation creates a reverse entry on top so the audit trail stays intact. Every cross-table write happens in a single database transaction (it all saves or none of it does). Every action passes a full validation pipeline (balance, period, account status, intercompany, FX consistency). Auditors who have looked at it have approved.

What if ERPClaw disappears?

You still have the software. open source license, your code, your data, your server. The GitHub repo at github.com/avansaber/erpclaw is forkable by anyone. Compare to the QuickBooks Desktop sunset Intuit announced in 2023, which forced thousands of businesses onto QBO whether they wanted to be there or not.

## Try it

Install takes five minutes. The whole ERP, on your machine.

`clawhub install erpclaw`
That puts the core ERP (full accounting, inventory, HR, payroll) on your server. From there, talk to it: “I run a 12-person retail shop in Portland, set me up.” It picks the industry vertical, installs the relevant modules, generates the chart of accounts, and is ready for transactions.

If you want to read first:

- /features for the full module breakdown

- /pricing (spoiler: it is free)

- /compare/quickbooks for the side by side

- /for/small-business for the SMB pitch

- /docs for install and the QuickBooks migration guide

Repo: github.com/avansaber/erpclaw. Star it if you want to follow along, we ship weekly. Email `support@avansaber.com` if you hit something broken.

QuickBooks is the right software for what it is. ERPClaw is the right software when you have outgrown what QuickBooks is. The line between those is closer than most founders realize. If you are tracking inventory in a spreadsheet, paying for Synder, running two QBO subscriptions, or hiring a bookkeeper to build reports the product cannot, you crossed it a while ago.

Time to install.

---

## Rillet Alternative: Free AI Accounting for Stripe
URL: https://www.erpclaw.ai/blog/rillet-alternative-erpclaw-stripe/

# Rillet Alternative: Free AI Accounting for Stripe
Rillet quotes run $2,000+/month (founder-reported). ERPClaw does AI-native Stripe accounting inside a full ERP for $0, open source.

There is a quiet pattern I keep running into when I talk to SaaS founders. They are doing $5K, $20K, $80K of MRR. They are on Stripe. Their books are wrong, and they know it. They have looked at Rillet, nodded along to the demo, asked the price, and gone back to a Google Sheet because the quote was bigger than their monthly revenue.

This post is for those founders, and for the people already paying Rillet who are starting to wonder whether they need to.

I want to do something most “Rillet alternative” posts do not. I want to say nice things about the competitor. Rillet is good software. The team is sharp. The dashboard is the best I have seen in this category. If you read this whole post and decide Rillet is the right call, I will have helped you make a better decision, not failed at writing it.

But I want to make a different argument. The primitives Rillet charges enterprise pricing for are commodity business logic now, and we are at the start of a category change in how SaaS founders buy accounting tools. ERPClaw is the wedge, and the structural difference that makes the wedge work is that ERPClaw is AI-native and Rillet is AI-decorated. Rillet shipped a polished web product designed for analysts who click through forms; the AI features sit on top. ERPClaw was architected from the first commit with the AI assistant as the primary interface, the spec as the source of truth, and the GL posting rules as auditable code. That is not something Rillet can retrofit without throwing the codebase away. The full argument is in AI-decorated vs AI-native software.

## What Rillet does brilliantly

Let me get this out of the way first, because I think most “alternative to” posts are dishonest about the competition.

Rillet has done four things really well.

The dashboard. Rillet’s UI is the gold standard for SaaS accounting tools right now. The close cycle, revenue waterfall, deferred revenue rollforward, cohort views. Dense without being cluttered. It is the kind of thing you only build with a real design team and a real budget.

ASC 606 revenue recognition. Their strongest area. ASC 606 is the rule that says you cannot book a year of subscription revenue the day a customer pays you upfront; you have to spread it across the months you actually deliver the service. Rillet implements the full five-step model: contracts, performance obligations, transaction price allocation, recognition schedules, journal entries. They handle modifications, variable consideration, and the edge cases auditors actually ask about.

Three-layer per-transaction reconciliation. Every Stripe charge, refund, fee, and adjustment becomes its own journal entry. Payouts reconcile against the bank statement. The Stripe Clearing account nets to zero across the period. This is the right pattern. Most tools either skip it or do daily summaries and pretend it is the same thing.

Continuous close. Rillet’s “zero-day close” pitch is real. Keep up with the workflow and you close the books on day one of the next month instead of dragging it out for ten days. For a controller at a $50M ARR SaaS company, that is meaningful.

If you are a Series B SaaS with a controller, a CFO, and budget to make this disappear from your worry list, Rillet is a reasonable buy.

## Why their pricing is what it is

Rillet’s published price is “talk to sales.” Founder-reported quotes land in the $2,000 to $10,000 per month range (the same range we track on the Rillet comparison page), plus a 45-day implementation engagement that runs into the tens of thousands. There is no public starter tier. There is no free version.

This is not because Rillet is greedy. It is the math of their business.

Rillet has raised over $108 million from Sequoia, a16z, and ICONIQ (as of June 2026). They are a well-funded team. Their architecture is multi-tenant SaaS, hosted on their infrastructure, with CPAs on staff doing implementations and a dedicated customer success motion. Under that cost structure, servicing a $10K MRR founder profitably is hard. CAC runs into the tens of thousands. Support load per customer is high. The math does not work at $300 a month.

So Rillet does the rational thing. They price for the segment that can absorb it. That segment is mid-to-late-stage SaaS. Postscript, their reference customer, is a $100M+ ARR business (per Rillet’s Series B announcement, August 2025). The pricing fits the customer, the customer fits the pricing. Founders at $5K, $20K, $80K MRR are not the problem with Rillet’s pricing. They are just not the customer.

## What changes when these primitives go open-source

Here is the part the market has not fully metabolized yet.

ASC 606 logic is not a trade secret. Three-layer reconciliation is not a trade secret. MRR, Connect platform fee accounting, deferred revenue rollforwards. None of this is unique IP. It is accounting rules published since 2014 and double-entry patterns published since the 1490s. The only reason it has been expensive is that implementing it correctly required a team of CPAs and engineers working together for a year. That cost made it look proprietary.

That cost is collapsing. An ERP consultant who can write the spec, plus an LLM that can implement the spec, can produce the same primitives in weeks instead of years. I am not speculating. I have done it. ERPClaw’s Stripe integration covers the same primitives Rillet covers: payments, payouts, fees, disputes, subscriptions, ASC 606 revenue recognition, and Connect platform fees. The whole thing is open source. You can read the GL posting rules at /docs/stripe/asc-606/. You can audit the math. You can fork it.

When the implementation cost goes from “a Series A round” to “a focused spec-writing engagement,” the pricing model that depended on that cost no longer makes sense for the segment that cannot afford it. That is the wedge. Not “Rillet is bad and we are good.” Rillet is fine. The wedge is that the price of these primitives just went to zero, and someone has to be the company that ships them at zero. We are choosing to be that company.

## ERPClaw vs. Rillet, feature by feature

Here is the comparison table from /compare/rillet, expanded with notes.

| Feature
| ERPClaw
| Rillet

| Price
| Free (open source license)
| $2,000 to $10,000/mo (founder-reported quotes), enterprise sales

| Implementation
| 5 minutes via CLI
| 45 days with CPA team

| Target market
| Any US SMB
| VC-funded SaaS startups

| Stripe scope
| Full coverage
| Full coverage

| ASC 606 revenue recognition
| Yes, five-step model
| Yes, five-step model

| Per-transaction posting
| Yes, every charge/fee/refund
| Yes

| Three-layer payout reconciliation
| Yes
| Yes

| Connect platform fees
| Yes (direct, destination, separate)
| Yes

| MRR / ARR / churn metrics
| Yes
| Yes

| Manufacturing / BOM / MRP
| Yes (full)
| No

| Inventory (batches, serials, ledger)
| Yes
| No

| HR and payroll
| Yes (built-in)
| No (integrates Gusto/Rippling)

| CRM
| Yes (built-in)
| No (integrates Salesforce/HubSpot)

| Industry verticals
| 14 (health, edu, legal, retail, etc.)
| SaaS-only

| Interface
| Natural language chat (Telegram/web)
| Web application

| Deployment
| Self-hosted, your server
| Cloud-only SaaS

| Database
| SQLite or PostgreSQL via PyPika, your machine
| Proprietary cloud

| Architecture
| AI-native (assistant is the primary interface)
| AI-decorated (web UI with chat features added)

| Open source
| Yes (open source)
| No (proprietary)

| API
| Deep, scriptable JSON API across the whole system
| No public API

| Self-improving (gap detection, module gen)
| Yes (ERPClaw OS)
| No

The columns where Rillet wins are the columns where it should win. They are an enterprise SaaS product built for one market. The columns where ERPClaw wins are the columns where being open-source and horizontally scoped pays off.

## Where Rillet still wins

I want to be specific here, because I do not want to sell you something that is not right for you.

UI polish. Rillet’s web UI is genuinely better than anything you will see in webclaw today. We are catching up. If you need a web dashboard right now, today, that you can put a junior controller in front of and have them be productive in an hour, Rillet wins. ERPClaw gets you there through chat, which has its own advantages but is a different muscle.

Support tier. Rillet has CPAs on staff. If you need someone to walk you through a quirky Connect platform structure or a complicated multi-entity consolidation, you can pay them and get a human who knows the answer. ERPClaw has me, a Discord, and the source code. That is fine for engineers and curious founders. It is not fine for a CFO who needs SLA-backed support.

Established brand. When a Big 4 auditor walks into a portfolio review and asks what your accounting stack is, “Rillet” lands differently than “an open-source project I installed last month.” That brand reputation is a real thing. We are working on it. We do not have it yet.

SOC 2 and compliance certifications. Rillet has the certifications procurement teams want to see. ERPClaw is self-hosted, so compliance is mostly about your infrastructure, not ours. But if your buyer wants SOC 2 attestation on the vendor itself, Rillet has it and we do not.

If any of those four are dealbreakers, go book the Rillet demo. The point of this post is not to talk you out of a tool that fits.

## Where ERPClaw wins decisively

Here is where the open-source posture is not just a price difference, it is a structural one.

Price. ERPClaw is zero dollars. Forever. Not a trial, not a freemium. Open source, you own the code, you run it on your own infrastructure. For a founder at $20K MRR, this is the difference between “I cannot afford proper books” and “my books are correct.”

Source code. You can read every GL posting rule, audit the ASC 606 logic, and fork the repo if you do not like a decision we made. Rillet’s GL rules are proprietary. If their engine miscategorizes a Connect application fee, you cannot see why.

Self-hosted, your data. ERPClaw runs on your machine. Your Stripe data syncs into a SQLite or PostgreSQL database on your server (PyPika abstracts the layer, so the same code runs on either). If we disappear tomorrow, your books keep working. With Rillet, your data is in their cloud, and the exit is a CSV dump.

No implementation cost. Rillet’s 45-day implementation often runs $20K to $50K. ERPClaw installs in five minutes via the CLI.

Full ERP scope. Rillet is a finance tool. ERPClaw covers the full back office: manufacturing, inventory, HR, payroll, CRM, projects, and 14 vertical industry packs. If you ever pivot, expand, or acquire into a non-SaaS business, ERPClaw scales with you. Rillet does not.

## When you should pick Rillet anyway

Let me write the section that competitor comparison posts usually skip.

You should pick Rillet if:

- You are a Series B+ SaaS company with a controller, and the cost of the tool is a rounding error against the cost of the controller.

- You need SOC 2 attestation on the vendor itself, not on your infrastructure.

- You want a CPA-led implementation and ongoing white-glove support.

- You need the polished web UI today and cannot wait for webclaw to catch up.

- You have an auditor who has already approved Rillet for your portfolio and switching tools introduces audit risk.

- You genuinely do not care about owning the source code or the data, and the cloud convenience is worth more than the lock-in.

If two or more of those apply, Rillet is a sensible buy. Send me a note and I will tell you the same thing in person.

You should pick ERPClaw if:

- You are a founder, solo or small team, and the price is the blocker.

- You want to own your data on your own infrastructure.

- You want to read the GL posting rules and audit them yourself.

- You need more than just accounting (inventory, HR, manufacturing).

- You are comfortable with a chat-first interface and a thinner web UI.

- You believe the open-source posture is a long-term hedge against vendor risk.

## How to migrate from Rillet to ERPClaw

If you are on Rillet today and want to try ERPClaw, here is the path, short version:

Step 1: Install ERPClaw. On a VPS, on your laptop, anywhere with Python 3.10+. The install walkthrough at /docs/stripe/install-walkthrough takes about five minutes.

Step 2: Connect Stripe. Either via the Stripe Marketplace (two-click OAuth) or via a restricted API key for air-gapped installs. ERPClaw will pull your full Stripe history.

Step 3: Run in parallel for a month. Do not cut over yet. Let ERPClaw and Rillet both post journal entries for a full close cycle. Compare the GL outputs. They should match. Deltas are usually about chart-of-accounts mapping, not engine bugs.

Step 4: Reconcile. The three-layer reconciliation in ERPClaw should net to zero on the Stripe Clearing account. If Rillet’s nets to zero too and the totals match, you have validated the engine.

Step 5: Cut over. When your auditor or CFO is comfortable, cancel Rillet. Export your Rillet history as a CSV for archival in case you need it for prior-year audits.

Step 6: Optional. If you want managed hosting, ERPClaw Cloud Managed launches in Q4 2026. Same source code, you just do not run the server. Pricing will be a fraction of what Rillet charges.

## FAQ

Does ERPClaw handle ASC 606 as well as Rillet?

For the standard SaaS cases, yes. We implement the five-step model across `advacct_revenue_contract`, `advacct_performance_obligation`, `advacct_variable_consideration`, and `advacct_revenue_schedule` tables. ERPClaw handles straight-line monthly recognition, contract modification, and the cancellation cases. The full breakdown is at /docs/stripe/asc-606/.

The places where Rillet still has an edge: relative SSP allocation across bundled performance obligations (we currently use SSP directly as allocated price 1:1, which is fine for single-obligation SaaS but less precise for bundles), and complex variable consideration with constraint adjustments. If your contracts are simple “$X per month for service Y,” ERPClaw is fully sufficient. If your contracts are bundles with implementation fees, support tiers, and usage-based components requiring sophisticated allocation, Rillet’s engine is more battle-tested.

Does ERPClaw compute MRR the same way Rillet does?

Yes, from your actual Stripe subscription data, including expansion, contraction, downgrades, pauses, and resumes. Detail at /docs/stripe/mrr-reports/. The math is parsed from Stripe subscription items directly, not estimated from invoice totals.

Does ERPClaw handle Stripe Connect platform fees?

Yes. ERPClaw handles direct charges with application fees, destination charges, and separate charges with transfers. Each posts the right journal entries to the platform’s books and the connected account’s books. Detail at /docs/stripe/connect-platform-fees/.

What about refunds and disputes?

Refunds post as reversal entries in the GL. Disputes hold funds in a disputes-in-progress account until resolution, then either reverse the original charge (if lost) or release the held funds (if won). The pattern matches what Rillet does.

How does the three-layer reconciliation work?

Layer 1: every Stripe transaction becomes a journal entry. Layer 2: every Stripe payout becomes a bank deposit, matched to the bank statement. Layer 3: the Stripe Clearing GL account nets to zero across the period. If it does not zero, ERPClaw flags it. Same pattern as Rillet, same pattern as A2X.

Is ERPClaw audit-ready?

Yes for the engine. The GL is immutable (cancel = reverse, never edit), every entry has a complete audit trail, and the 12-step GL validation runs on every posting. What ERPClaw does not give you is a vendor SOC 2 report, because there is no vendor. You are the vendor of your own infrastructure. If your auditor wants vendor attestation rather than infrastructure attestation, Rillet has that and we do not.

What if I outgrow ERPClaw?

You migrate. The data is yours, the schema is documented, and you can export to any standard accounting tool. We designed against vendor lock-in. If Rillet is a better fit in two years, switch to Rillet then.

## The honest closing

Rillet is a good company building a good product. I respect what they have built. If they read this post, I hope they take it as the backhanded compliment it partly is.

But they are pricing for a market that is shrinking as a fraction of the addressable founder population. Every month, more SaaS founders start up, and the share who can absorb founder-reported quote levels ($2K to $10K monthly) for accounting tooling is going down. The category is bifurcating. A thin slice at the top will keep paying enterprise SaaS prices for white-glove service. Everyone else needs the same primitives at zero cost.

ERPClaw is built for everyone else. The primitives are the same. The polish, brand, and white-glove support are not, and may never be. That is fine. We are not trying to be Rillet at a discount. We are trying to be the open-source, AI-native standard for what these primitives should look like once the implementation cost collapses.

Start at /features/stripe. Pricing is at /pricing. The full side-by-side is at /compare/rillet.

Install in five minutes. Sync your Stripe data. Run it in parallel with whatever you have today. Tell me if the journal entries match. They should.

---

## The Sage Intacct Alternative That Costs $0
URL: https://www.erpclaw.ai/blog/sage-intacct-alternative-free-erp/

# The Sage Intacct Alternative That Costs $0
Sage Intacct pricing is quote-gated and climbs with modules. ERPClaw posts a real general ledger, covers the whole business, and costs $0 forever.

There is a line item on your software bill that grows every year and nobody on the finance team flags it. It is the Sage Intacct subscription. It started around $15,000 a year for the entry tier with three modules and two entities. Then you added Fixed Assets. Then another entity for the second LLC. Then the per-user count crept up because Ops needed read access. Then the renewal landed and the year-one discount quietly evaporated. You are now between $25,000 and $60,000 a year, you paid an implementation partner another $40,000 two years ago, and the renewal email lands like clockwork in Q3.

This post is for the controller, the CFO, or the finance manager who looks at that number and asks the question nobody on the Sage account team wants you to ask: is there a real alternative now, given everything that has changed about how software gets built?

The honest answer is yes, with caveats, and the caveats matter as much as the yes.

ERPClaw is an open-source AI-native ERP. It is open source, self-hosted, database-agnostic via PyPika, and it implements the primitives mid-market finance leaders care about: immutable double-entry GL, multi-entity, ASC 606 revenue recognition for SaaS, audit-ready trail with chain hashing, and a posting validation pipeline that runs every transaction. It is the only AI-native option in this category. Every other comparable product, Sage Intacct included, is AI-decorated: a chatbot was added to a 1990s general ledger product as a paid module. ERPClaw was designed AI-native from the spec on day one.

The wedge is not “we are cheaper.” The wedge is “same primitives, structurally 100x cheaper, with an architecture mid-market vendors cannot replicate without a full rewrite.”

## What Sage Intacct does well

Sage Intacct is the best product in its tier and pretending otherwise is dishonest.

Multi-entity consolidation that auditors love. Sage Intacct was built on a true multi-entity data model from the start. Inter-entity eliminations, currency translation per entity, ownership percentages, partial consolidations, all of it works the way a Big Four auditor expects. If you have eight entities and you close in five business days, this is part of the reason.

Multi-currency depth. Functional currency at the entity, reporting currency at the consolidated parent, translation methods configurable per account, gains and losses booked correctly to OCI vs P&L per FAS 52. It is not a feature you turn on. It is woven into the data model.

ASC 606 contract management. The Contracts and Revenue Management modules handle multi-element arrangements, performance obligation tracking, standalone selling price allocation, and the deferred revenue waterfalls SaaS billing produces. The product was built when ASC 606 went live and the implementation reflects that.

Established mid-market motion. When you say “we run on Intacct” to your auditor, your bank, or your investors, they know what you mean. You can hire an Intacct admin off LinkedIn in a week.

Premium support tier. Real account managers, real SLAs, real escalation paths. When the close is jammed at 11pm on day five, somebody picks up the phone.

The question is whether all of that is worth $25,000 to $60,000 a year plus a six-figure implementation, given that the primitives underneath are now buildable in open source with a small team and a code-generation feedback loop.

## The Intacct line item, fully loaded

Partner-quoted pricing ranges as of April 2026 (Sage Intacct does not publish list pricing; contact Sage or a certified partner for a current quote), for a typical 5-entity, 20-user mid-market install:

- Base Intacct (GL, AP, AR, Cash): $9,000 to $15,000 per year

- Multi-entity module: $4,000 to $8,000 per year

- Fixed Assets: $3,000 to $5,000 per year

- Contracts and Revenue Management (ASC 606): $6,000 to $12,000 per year

- Time and Expense: $3,000 to $5,000 per year

- Per-user fees, 20 users: $4,000 to $10,000 per year

- Implementation partner, year one: $40,000 to $120,000 one time

- Annual maintenance and re-implementation work: $10,000 to $30,000 per year

- Premium support tier: $5,000 to $15,000 per year

Steady-state year two and beyond, $35,000 to $75,000 per year plus one-time setup. Over five years most mid-market companies on Intacct spend $250,000 to $500,000 on the financial system alone.

This is not a complaint. The product is priced fairly given Sage’s cost structure: thousands of employees, a sales force, an implementation partner channel that takes a cut, the same SaaS unit economics every mid-market vendor faces.

ERPClaw does not have any of that cost base. Built at AvanSaber Inc by a small co-founder team using Claude Code with a spec-first methodology, the marginal cost of adding a customer is approximately zero. No sales team, no per-seat license, no implementation partner channel, no quarterly renewal forecast to defend.

## Where ERPClaw matches the primitives

The skeptical mid-market controller’s first reaction to “open source ERP” is usually a flashback to a half-finished ERPNext install where the GL did not actually balance. Fair instinct. Be specific about what ERPClaw implements that maps to the Sage Intacct primitives that matter.

Immutable GL with chain hashing. Every posted journal entry carries a SHA-256 hash that includes the previous entry’s hash. Tampering breaks the chain. The `gl_entry` table has no `updated_at` column by design. Cancellation produces a reversing entry, never an edit.

Posting validation that cannot be skipped. Every GL posting runs through a validation pipeline before commit: balance, period status, account status, intercompany consistency, FX consistency, dimension validity, sign convention, period lock, currency code, exchange rate presence, posting account type, submission state. Any failure rolls back the whole transaction. No partial posts.

Multi-entity via shared database. ERPClaw’s data model carries entity at the row level, not at the database level. One install supports N entities with intercompany eliminations and consolidated reporting in the parent currency.

ASC 606 for SaaS via Stripe. The Stripe ASC 606 engine handles the five-step model end to end for subscription revenue. Deferred revenue waterfalls, contract modifications, and revenue catch-ups are in the box. For pure-play SaaS on Stripe, this is the most-asked Intacct module and ERPClaw ships it natively.

Database-agnostic. ERPClaw’s data layer is built on PyPika, a Python query builder that compiles to ANSI SQL. Default install is SQLite for simplicity, but the same code runs against Postgres or MySQL. You are not locked into a single engine, and you are not locked into our hosted cloud.

Audit-ready trail. Every action writes to an immutable audit log with user, timestamp, action, before-state, after-state, and chain hash. Auditors who have walked through it have signed off.

## Where Sage Intacct still wins

Same honesty in the other direction.

Full multi-currency consolidation per FAS 52. ERPClaw v1 supports multiple currencies for transactions, but the full functional-vs-reporting currency consolidation with current-rate vs historical-rate translation per account, OCI handling for non-functional gains, and the cumulative translation adjustment account is partial. If you have entities in five currencies and your auditor checks CTA every quarter, Sage is more complete today.

Established sales motion and ecosystem. You cannot hire an ERPClaw admin off LinkedIn this week. There are no Big Four implementation partners with a dedicated practice. If your CFO’s risk profile requires “vendor with a 20-year track record and a Gartner Magic Quadrant position,” ERPClaw is not that. We are 18 months old.

Premium support tier with SLAs. ERPClaw support today is GitHub issues, email, and the founder. There is no 24/7 phone line, no named account manager. We will offer an enterprise support tier (probably Q1 2027), but for now it is community plus email.

Pre-built reporting library. Sage Intacct ships hundreds of pre-built reports with mid-market polish. ERPClaw ships the standards (P&L, balance sheet, trial balance, cash flow, AR aging, AP aging, GL detail) and a query layer you can extend. The breadth of out-of-the-box dashboards is still building.

Industry-specific compliance modules. Sage has dedicated modules for nonprofit fund accounting (FASB 117), construction job costing with WIP schedules, and healthcare grant management that are genuinely deep. ERPClaw has equivalents in our vertical modules, but the depth in some industries is still building out.

If any of those five matter more to you than $250,000 over five years, stay on Intacct. The wrong move is switching for the savings and discovering at year-end that the close takes nine days instead of five.

## The AI-native plus open-source structural argument

Here is the part that does not show up on the comparison table.

Sage Intacct cannot become AI-native. They can add AI features (and they have, with Sage Copilot), but those features are decorations on a code base designed in the 1990s for a forms-and-workflow paradigm. Genuinely AI-native means the system was specified, built, and shipped with the assumption that the primary interface is natural language and the primary author of new code is a model. That assumption changes the architecture top to bottom: the action layer, the schema design, the test suite, the documentation. You cannot retrofit it without a full rewrite, and a public mid-market vendor with thousands of paying customers cannot afford the years-long disruption.

ERPClaw was built AI-native from day one. Every action is specced before it is coded. Every action has a contract test the model can read and verify. Every module ships with a SKILL.md that lets any AI assistant operate the system in natural language. The full argument is in AI-decorated vs AI-native software.

The compounding effect is what matters. Today, ERPClaw is roughly at parity with Sage Intacct on the core primitives, with gaps on the edges I listed. In 18 months the gap closes (fully OCI-aware multi-currency, broader pre-built reports, more vertical depth) because the build velocity of an AI-native open-source project compounds in ways a SaaS vendor’s roadmap cannot. Meanwhile Sage’s pricing keeps creeping. Year five Intacct is more expensive than year one. Year five ERPClaw is still $0.

This is what I mean by structural. It is not just that we are cheaper. The cost structure of an AI-native open-source ERP cannot be matched by an incumbent without burning the boats.

## Side by side

The full table is at /compare/sage-intacct. Condensed:

| Feature
| ERPClaw
| Sage Intacct

| License cost
| Free (open source)
| $15,000+ to $60,000+/year

| Implementation cost
| $0 (5-minute install)
| $40,000 to $120,000

| Implementation time
| Minutes
| 3 to 12 months

| Multi-entity
| Yes (shared DB)
| Yes (mature)

| Multi-currency
| Partial v1
| Yes (full FAS 52)

| ASC 606 (SaaS)
| Yes (Stripe-native)
| Yes (Contracts module)

| Inventory + Manufacturing
| Yes
| Yes

| Immutable GL
| Yes (SHA-256 chain)
| Yes

| AI capability
| AI-native (architecture-level)
| AI-decorated (Copilot module)

| Database
| PyPika, any SQL backend
| Sage cloud only

| Self-hosted
| Yes
| No

| Open source
| Yes (open source)
| Proprietary

| Support
| Community + email
| Premium tier with SLA

## When to stay on Sage Intacct

Honest “do not switch” cases:

You already paid for the implementation and the team is trained. If you spent $80,000 to roll out Intacct two years ago and your accounting team is fluent in it, the switching cost is real and the savings take years to recover. Wait for the next platform decision (acquisition, re-architecture, CFO transition).

Multi-currency consolidation is critical right now. If you are closing in three currencies tomorrow with full FAS 52 OCI handling, Sage is more complete today. Wait six months for ERPClaw v2 multi-currency.

Your industry needs Sage’s vertical depth. Specifically: large nonprofits with deep fund accounting, construction firms with WIP schedules and percent-complete revenue, or healthcare with grant compliance.

You need an SLA in writing for compliance reasons. SOX-eligible public companies and certain regulated industries cannot run their books on community-supported software, full stop.

If none of those apply, the math is hard to argue against. For the comparison to a different mid-market option, see /compare/netsuite.

## How a migration actually works

Scoping version. Full guide publishes Q3 2026.

-
Export from Intacct. Customers, vendors, items, employees, chart of accounts, dimensions (departments, locations, classes), GL detail. CSV across all of these. Half a day.

-
Map your chart of accounts. ERPClaw ships a US default chart of accounts as a starting point. The mapping script reads your Intacct export, matches by name and type, surfaces unmatched accounts. One full day with the controller in the room.

-
Map your dimensions. Intacct dimensions (department, location, class, project) map one-to-one onto ERPClaw’s model.

-
Pick a cutover date. Standard practice: start of a fiscal quarter. Close Intacct cleanly, freeze it, then open ERPClaw with opening balances as of day one of the new quarter.

-
Import opening balances. A single command loads trial balance, AR aging, AP aging, and inventory valuation. The posting validation rejects any imbalance before commit.

-
Run parallel for one quarter. Do not skip this. Post the same transactions to both systems for one full quarter and reconcile at month-end and quarter-end. Only kill Intacct when the books match to the cent.

For a typical 5-entity, 20-user deployment, the full migration is six to ten weeks of part-time work with one accountant lead and one technical lead. Compare to the three-to-twelve-month timeline most companies experienced when they first deployed Intacct.

## FAQ

Is ERPClaw really audit-ready for a mid-market financial audit?

Yes for the GL and trial balance side. Auditors who have walked through the immutable GL, chain hash audit trail, posting validation, and per-action audit log have approved it for substantive testing. Pieces still in progress for full SOX-eligible attestation: SOC 2 (Q3 2026) and certain ITGC controls around hosted deployments. If you self-host, you own your own ITGC environment, the same as any on-prem ERP.

How does ASC 606 work compared to Sage’s Contracts module?

Built natively for SaaS revenue and fully integrated with Stripe. Multi-element arrangements, performance obligation tracking, standalone selling price allocation, contract modifications, and revenue catch-ups are in the engine. For pure SaaS on Stripe billing, our implementation is at parity. For complex contracts with non-Stripe billing rails, Sage’s Contracts module is broader today.

Can I run ERPClaw against Postgres instead of SQLite?

Yes. The data layer is built on PyPika, which compiles to ANSI SQL. Default install uses SQLite for simplicity, but production deployments commonly run Postgres. Full deployment doc at /docs.

Does ERPClaw support multi-entity intercompany eliminations?

Yes for the standard cases. Intercompany payables and receivables eliminate at the consolidation level, intercompany loans net out, and the consolidated trial balance reflects the eliminations. For complex partial-ownership consolidations and minority interest accounting, the v2 roadmap closes the remaining gaps.

What is the support model if something breaks during close week?

GitHub issues, email to support@avansaber.com, and the founders are reachable directly. The source is open, so worst-case you (or a contractor) can find the issue. That is not the same as a Sage account manager with a phone number. If an SLA is non-negotiable, stay on Sage until our enterprise tier launches.

What if ERPClaw the project disappears?

You still have the software, the source, the database, and the right to fork it forever (open source). Compare to a SaaS vendor that gets acquired, sunsets a product, or raises prices on you mid-contract.

## Try it before you renew

The next Sage Intacct renewal is probably six to twelve months out. That is enough time to install ERPClaw on a side server, load a single legal entity’s prior-year trial balance, run parallel for a month, and decide whether the gap matters for your situation.

`clawhub install erpclaw`
That installs the full ERP on your server in about five minutes. From there, talk to it: “load chart of accounts and trial balance from this Intacct export, generate me a P&L for last quarter.” Or use the web interface, or the Telegram bot, or call the action API directly.

Read first:

- /compare/sage-intacct for the side-by-side

- /features for the full module breakdown

- /pricing (it stays free, open source)

- /docs/stripe/asc-606 for the SaaS revenue engine

- /compare/netsuite if you are also evaluating NetSuite

- /blog/ai-decorated-vs-ai-native-software for the architecture argument

For the original product, see Sage Intacct. It is a good product. The question this post asks is not “is Intacct bad” (it isn’t), but “is the open-source AI-native alternative good enough that the renewal price stops making sense.” For a growing share of mid-market finance leaders, the answer in 2026 is yes.

Renewal letter coming in Q3? Install ERPClaw this weekend. You will know within a month whether it covers what you need. If it does, you have months of runway to migrate properly. If it does not, you have learned what gaps to push Sage on at the renewal table.

Either way, you walk into that meeting with negotiating room you did not have a year ago.

---

## Sage Intacct Alternatives 2026: the Open-Source Take
URL: https://www.erpclaw.ai/blog/sage-intacct-alternatives-2026/

# Sage Intacct Alternatives 2026: the Open-Source Take
A vendor-neutral guide to the real Sage Intacct alternatives in 2026, plus a simple test for telling genuine AI-native finance software from AI bolted on.

Short answer. People leave Sage Intacct for one of three reasons. They want broader ERP, not just financials (look at NetSuite or Acumatica). They want finance software built around AI from the start (look at the AI-native tools like Rillet or DualEntry, and at ERPClaw). Or they want to stop renting and own their system outright (open source: ERPNext, Odoo, or ERPClaw). This guide covers all three honestly, and gives you a simple test to see through the “AI” label every vendor now wears.

Sage Intacct earned its place. It is genuinely strong cloud financial management: a solid general ledger, real multi-entity consolidation, and reporting that finance teams like. The reason people start searching is usually the renewal. The entry tier sits around $15,000 a year, then modules, entities, and user counts push a typical mid-market bill toward $25,000 to $60,000 a year, plus the implementation you paid a partner for. When that number grows every year and the AI story is a chat box added to a 1990s ledger, the question gets asked.

Here are the real alternatives, grouped by what you are actually trying to do.

## The test: genuine AI-native, or AI bolted on?

Every finance product now says “AI.” Most of it is a summarize button on software designed decades ago. Before you shortlist anything, run three questions:

- Does the AI do the work, or describe it? A summary of a report is not an agent that posts the journal entries and reconciles the accounts. Ask to see an action completed, not narrated.

- Was the system built around AI, or retrofitted? Adding a model to an old codebase keeps the old cost base and the old workflows. Software specced around AI from the start behaves differently.

- Can you talk to it to get something done? “Reclass last month’s AWS spend to R&D and re-run the department P&L” should just happen. If the only interface is forms with a chat sidebar, that is AI-decorated, not AI-native. The full argument is in AI-decorated vs AI-native software.

Hold each option below to that test.

## Group 1: broader ERP (you outgrew financials-only)

Sage Intacct is finance-first. If you need inventory, manufacturing, or a deep app ecosystem alongside accounting, these are the usual moves. All paid, all AI added on rather than built in.

- NetSuite. The bigger incumbent. Broader than Intacct (inventory, manufacturing, a large app marketplace), and priced for it. See our NetSuite alternatives guide if it is also on your list.

- Acumatica. Cloud ERP with user-count-friendly pricing and broad functionality. AI bolted on.

- Microsoft Dynamics 365 Business Central. The natural choice if you live in the Microsoft stack. Copilot is the AI layer, added to an existing product.

- SAP Business One. Aimed at manufacturing and distribution rather than pure finance. Capable, older, AI as a side feature.

Pick one of these if the real gap is breadth, and AI is a nice-to-have.

## Group 2: the AI-native newcomers

The genuinely new category: finance systems specced around automation from the start. Mostly venture-funded, closed source, and paid.

- Rillet. An AI-native general ledger built for mid-market finance, designed by accountants. Strong on the automation story. Closed and paid.

- DualEntry. AI-native ERP aimed at finance teams scaling toward IPO. Also closed and paid.

- Campfire and similar. A growing set of well-funded entrants in the same space.

These pass the AI-native test. The trade-off is the familiar one: you rent the software, the source is closed, and your data lives in their cloud on their pricing roadmap.

## Group 3: open source (own your finance system)

If the renewal treadmill is the thing you want off, this is your group.

- ERPNext and Odoo. The established open-source options. Broad and genuinely yours to host. The catch against the test above: AI arrives through plugins, so they are open source but AI-decorated.

- ERPClaw. This is the one we build, so weigh it accordingly. It is the option that is open source and AI-native and free. The agent posts double-entry journals end to end rather than summarizing a screen. It covers the finance primitives Intacct buyers care about (immutable double-entry GL, multi-entity, ASC 606 revenue recognition, audit-ready trail, a validation pipeline on every posting) plus the rest of the back office, in one shared database. It is $0 forever, self-hosted on your own infrastructure. The honest gaps versus Intacct: the deepest statutory multi-entity consolidation, a large partner ecosystem, and a contractual support SLA. If those are dealbreakers, a paid option fits better, and we say so.

## Side by side

| Option
| AI
| Open source
| Self-host
| Price posture
| Best fit

| Sage Intacct
| Add-on
| No
| No
| Paid, grows yearly
| Finance-heavy mid-market

| NetSuite
| Add-on
| No
| No
| Six-figure
| Broad ERP, ecosystem needs

| Acumatica
| Add-on
| No
| No
| Paid
| Seat-growth companies

| Dynamics 365 BC
| Add-on
| No
| No
| Paid
| Microsoft shops

| Rillet
| Native
| No
| No
| Paid
| AI-first mid-market finance

| DualEntry
| Native
| No
| No
| Paid
| Finance teams scaling to IPO

| ERPNext / Odoo
| Plugin
| Yes
| Yes
| Free or paid tiers
| Open-source traditionalists

| ERPClaw
| Native
| Yes
| Yes
| $0 forever
| Open-source, AI-native, cost-led

The table is not meant to show ERPClaw winning every row. It is meant to show that open source and AI-native at the same time is a near-empty cell. The commercial roundups skip it because there is no affiliate revenue in a free tool. That gap is why this guide exists.

## Which group are you in?

- You need more than financials. NetSuite or Acumatica. The breadth is the reason to pay.

- You want AI to actually run the close, and you will pay for it. Shortlist Rillet or DualEntry, and put ERPClaw beside them as a free baseline to measure against.

- You want to stop renting your ledger. ERPNext, Odoo, or ERPClaw. If you also want the AI to do the work rather than describe it, ERPClaw is the one that is both.

## FAQ

What is the cheapest real alternative to Sage Intacct?

Open source is the floor. ERPNext and Odoo have free community editions; ERPClaw is $0 forever and self-hosted, with AI built in rather than added. “Cheapest” only counts if it covers the finance functions you use, so check those before comparing prices.

Does any alternative match Sage Intacct on multi-entity?

For everyday multi-entity (multiple companies, multiple currencies, consolidated reporting), several do, including ERPClaw. Where Intacct still leads is the deepest statutory per-jurisdiction consolidation. If you run that, weigh it carefully.

Are the AI-native finance tools real, or marketing?

Both exist. Use the three-question test: does the AI complete actions, was the system built around AI, can you talk to it to get work done. Rillet, DualEntry, and ERPClaw pass it. A summarize button on a legacy ledger does not.

Can an open-source system give my auditors what they need?

Yes, when the GL is immutable, every posting passes validation, and the audit trail is intact. ERPClaw is built that way. Confirm audit-readiness with any open-source option before you commit.

How do I compare ERPClaw to Sage Intacct directly?

See ERPClaw vs Sage Intacct for the side by side, and the $0 Sage Intacct alternative for the detailed single-product case.

## Where to go next

- ERPClaw vs Sage Intacct, the detailed side by side

- The Sage Intacct alternative that costs $0

- NetSuite alternatives in 2026

- The 5 AI-native ERPs that actually earn the label

If you want to try the open-source AI-native option before the next renewal, install ERPClaw in a few minutes and run it alongside your current books for a period. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

Tagssage-intacctalternativesai-nativeerpopen-sourcecomparison
Share:XLinkedInHacker News

---

## SAP Business One Alternatives 2026: Open-Source Picks
URL: https://www.erpclaw.ai/blog/sap-business-one-alternatives-2026/

# SAP Business One Alternatives 2026: Open-Source Picks
A vendor-neutral guide to real SAP Business One alternatives in 2026, from Acumatica and NetSuite to open-source ERP, plus a test for AI-native software.

Short answer. People look for a SAP Business One alternative for one of three reasons. They want a modern cloud ERP without the reseller and implementation weight (look at Acumatica, NetSuite, or Microsoft Dynamics 365 Business Central). They want finance and operations software built around AI from the start, not a copilot added later (look at the AI-native tools, and at ERPClaw). Or they want to stop paying per-user licenses and own the system outright (open source: ERPNext, Odoo, or ERPClaw). This guide covers all three honestly, and gives you a simple test to see through the “AI” label every vendor now wears.

SAP Business One is a serious, proven product. SAP describes it as built for “small businesses and the lower midmarket,” a system that “can grow with them and meet their evolving needs,” and the install base backs that up: more than 83,000 customers and 1.2 million users across over 170 countries, served by roughly 850 partner organizations. It is strong in manufacturing and distribution, with real inventory, production, and multi-currency depth.

The reason people start searching is usually one of two things. The cost adds up, or the AI story does not match the SAP AI headlines. Here are the real alternatives, grouped by what you are actually trying to do.

## Why buyers leave: cost and the AI gap

Two forces push Business One buyers to look around.

The per-user license math. SAP does not publish one clean public price, and quotes vary by partner, region, and deal size, but the ranges are well documented. Cloud subscriptions land around 91 euros per user per month for the Professional license and 47 euros for Limited, while perpetual on-premise licenses run into the low thousands per user plus an annual maintenance fee. Independent pricing trackers put the effective cloud cost in the range of 110 to 219 dollars per user per month once modules and hosting are counted, before the partner-led implementation you pay separately. For a growing team, the per-seat model is the line item that climbs every year.

The AI gap. SAP’s headline AI story in 2026 is Joule, the agentic assistant SAP has pushed hard at Sapphire. It is real and it is deep, but it lives in the enterprise cloud tier. As of the SAP Business AI Q1 2026 release highlights, Joule and Joule Studio are generally available across S/4HANA Cloud, SuccessFactors, Ariba, and dozens of other enterprise solutions. SAP Business One, the SMB tier, is a different product on a different roadmap. If a reseller deck showed you Joule agents booking meetings and reconciling invoices, it is worth knowing that surface does not ship inside Business One. We walk through exactly what is embedded and what is not in SAP Joule and SAP Business One: what is embedded vs what is not.

## The test: genuine AI-native, or AI bolted on?

Every ERP now says “AI.” Most of it is a summarize button on software designed decades ago. Before you shortlist anything, run three questions:

- Does the AI do the work, or describe it? A summary of a report is not an agent that posts the journal entries and reconciles the accounts. Ask to see an action completed, not narrated.

- Was the system built around AI, or retrofitted? Adding a model to an old codebase keeps the old cost base and the old workflows. Software specced around AI from the start behaves differently.

- Can you talk to it to get something done? “Reconcile last month’s supplier payments and re-run the inventory valuation” should just happen. If the only interface is forms with a chat sidebar, that is AI-decorated, not AI-native. The full argument is in AI-decorated vs AI-native software.

Hold each option below to that test.

## Group 1: modern cloud ERP (you want breadth without the weight)

Business One is capable but traditional, and it usually arrives through a regional reseller with a project attached. If you want comparable breadth in a more modern cloud package, these are the usual moves. All paid, all with AI added on rather than built in.

- Acumatica. Cloud ERP with resource-based rather than per-user pricing, which appeals to teams that would otherwise watch a per-seat bill climb. Broad functionality across finance, distribution, and manufacturing. AI is a bolt-on layer.

- NetSuite. The bigger incumbent, broader and pricier, with a large app marketplace. If it is also on your list, our NetSuite alternatives guide covers it in the same honest frame.

- Microsoft Dynamics 365 Business Central. The natural choice if you already live in the Microsoft stack. Copilot is the AI layer, added to an existing product.

Pick one of these if the real gap is a modern cloud footprint and a wider ecosystem, and AI is a nice-to-have rather than the point.

## Group 2: the AI-native newcomers

The genuinely new category is finance and operations systems specced around automation from the start. Most are venture-funded, closed source, and paid.

- Rillet. An AI-native general ledger built for mid-market finance, designed by accountants. Strong on the automation story. Closed and paid.

- DualEntry. AI-native ERP aimed at finance teams scaling toward IPO. Also closed and paid.

- Campfire and similar. A growing set of well-funded entrants in the same space.

These pass the AI-native test. The trade-off is the familiar one: you rent the software, the source is closed, and your data lives in their cloud on their pricing roadmap.

## Group 3: open source (own your system, stop counting seats)

If the per-user treadmill is the thing you want off, this is your group.

- ERPNext and Odoo. The established open-source ERPs, both strong in the manufacturing and distribution territory Business One serves, and both genuinely yours to host. The catch against the test above: AI arrives through plugins, so they are open source but AI-decorated.

- ERPClaw. This is the one we build, so weigh it accordingly. It is the option that is open source and AI-native and free. The agent posts double-entry journals end to end rather than summarizing a screen, and you talk to it in plain language to run the books. It covers the finance primitives an SMB cares about (immutable double-entry general ledger, multi-entity, multi-currency, ASC 606 revenue recognition, an audit-ready trail, and a validation pipeline on every posting) in one shared database. It is $0 forever, self-hosted on your own infrastructure. The honest gaps versus Business One: deep manufacturing and production planning (Bill of Materials, MRP), a large certified-partner network, and a contractual support SLA. If those are dealbreakers, a paid option fits better, and we say so. The architecture case is laid out in ERPClaw vs SAP Business One: AI-native architecture.

## Side by side

| Option
| AI
| Open source
| Self-host
| Price posture
| Best fit

| SAP Business One
| Add-on
| No
| Optional
| Per-user, grows yearly
| Manufacturing and distribution SMB

| Acumatica
| Add-on
| No
| No
| Resource-based
| Seat-growth companies

| NetSuite
| Add-on
| No
| No
| Six-figure
| Broad ERP, ecosystem needs

| Dynamics 365 BC
| Add-on
| No
| No
| Per-user
| Microsoft shops

| Rillet
| Native
| No
| No
| Paid
| AI-first mid-market finance

| DualEntry
| Native
| No
| No
| Paid
| Finance teams scaling to IPO

| ERPNext / Odoo
| Plugin
| Yes
| Yes
| Free or paid tiers
| Open-source traditionalists

| ERPClaw
| Native
| Yes
| Yes
| $0 forever
| Open-source, AI-native, cost-led

The table is not meant to show ERPClaw winning every row. It is meant to show that open source and AI-native at the same time is a near-empty cell. The commercial roundups skip that cell because there is no affiliate revenue in a free tool. That gap is why this guide exists. For the deeper explainer on what “built around AI” actually means, see AI-native ERP.

## Which group are you in?

- You want a modern cloud ERP with breadth. Acumatica or NetSuite. The wider footprint is the reason to pay.

- You want AI to actually run the work, and you will pay for it. Shortlist Rillet or DualEntry, and put ERPClaw beside them as a free baseline to measure against.

- You want to stop renting per seat. ERPNext, Odoo, or ERPClaw. If you also want the AI to do the work rather than describe it, ERPClaw is the one that is both.

## What SAP itself says about the tier

It is worth grounding this in SAP’s own words rather than a competitor’s spin. SAP positions Business One squarely at the small-business and lower-midmarket buyer, an ERP that can “grow with them.” That is an honest description, and it is also the boundary. The autonomous-agent AI SAP markets under the Joule banner is aimed at the enterprise cloud tier, not this one, per SAP’s own Q1 2026 release highlights. So the practical question for a Business One buyer in 2026 is not “SAP or not SAP.” It is “do I want AI as a future add-on to a traditional SMB ERP, or do I want a system where the AI is the way you run it today.”

## FAQ

What is the cheapest real alternative to SAP Business One?

Open source is the floor. ERPNext and Odoo have free community editions; ERPClaw is $0 forever and self-hosted, with AI built in rather than added. “Cheapest” only counts if it covers the functions you use, so map your must-haves (inventory, manufacturing, multi-entity) before comparing prices.

Does any alternative match SAP Business One on manufacturing?

For core inventory, purchasing, and multi-currency, several do, including ERPNext, Odoo, and ERPClaw. Where Business One still leads is deep production planning (Bill of Materials, MRP, shop-floor control). If you run a real factory on it, weigh that carefully and test the alternative against your actual production flow.

Does SAP Business One include Joule AI?

Not as of 2026. Joule and its agentic features are generally available across SAP’s enterprise cloud products, not the Business One SMB tier. We break down exactly what is embedded and what is not in SAP Joule and SAP Business One.

Are the AI-native finance tools real, or marketing?

Both exist. Use the three-question test: does the AI complete actions, was the system built around AI, can you talk to it to get work done. Rillet, DualEntry, and ERPClaw pass it. A summarize button on a legacy ledger does not.

Can an open-source system give my auditors what they need?

Yes, when the general ledger is immutable, every posting passes validation, and the audit trail is intact. ERPClaw is built that way. Confirm audit-readiness with any open-source option before you commit.

How do I compare ERPClaw to SAP Business One directly?

See ERPClaw vs SAP Business One for the side by side, and the AI-native architecture comparison for how the two differ under the hood.

## Where to go next

- ERPClaw vs SAP Business One, the detailed side by side

- SAP Joule and SAP Business One: embedded vs not

- Sage Intacct alternatives in 2026

- NetSuite alternatives in 2026

If you want to see the open-source, AI-native option run your books before your next renewal, book a quick demo with a co-founder, or install ERPClaw and run it alongside your current system for a period. Repo at github.com/avansaber/erpclaw. Questions to `support@avansaber.com`.

## Sources

- SAP News, “10 New Customers a Day: The Global Success of SAP Business One” (customer, user, country, and partner counts; target-market framing)

- SAP News, “SAP Business AI: Release Highlights Q1 2026” (Joule and Joule Studio availability across enterprise cloud solutions)

- Business One Consultancy, “SAP Business One pricing for 2026” (list-price ranges for Professional and Limited licenses)

- CostBench, “SAP Business One Pricing 2026” (effective per-user cloud cost range)

Tagssap-business-onealternativesai-nativeerpopen-sourcecomparison
Share:XLinkedInHacker News

---

## SAP Joule and Business One: Embedded or Not?
URL: https://www.erpclaw.ai/blog/sap-joule-business-one-embedded-vs-not/

# SAP Joule and Business One: Embedded or Not?
SAP Joule is the headline AI inside S/4HANA Cloud, SuccessFactors, and Ariba. It is not inside SAP Business One. Here is what that means for an SMB founder in 2026.

A founder I spoke with last month had been pitched SAP Business One by a regional reseller. The deck had a slide about Joule, the same Joule that SAP demoed at Sapphire with autonomous agents booking meetings and reconciling invoices. She asked the question any reasonable buyer would ask: “so my Business One install gets that, right?”

It does not. Joule lives in the SAP enterprise tier. SAP Business One is the SMB tier. Different products, different roadmaps, different AI stories. The reseller deck was not lying, but it was reading a lot of weight onto a logo that does not belong to the product on offer.

This post walks through what Joule actually is in Q1 2026, which SAP products carry it, why it does not extend to SAP Business One, and what that means if you are sizing up Business One as a small business in 2026.

## What SAP Joule is in 2026

Joule is SAP’s generative AI copilot. It started in 2024 as a chat sidebar across a handful of SAP products. By Q1 2026 it has grown into something more ambitious. The numbers are easy to find in the SAP Q1 2026 release notes:

- Joule Studio is generally available, which lets customers and partners build their own agents on top of SAP data.

- A growing set of prebuilt agents ships across finance, supply chain, HR, sourcing, and customer experience.

- Over 2,400 Joule skills are available for agents to compose (as of June 2026).

- Two SAP-built foundation models, SAP-ABAP-1 and SAP-RPT-1, power the code generation and reporting paths.

At Sapphire 2026, SAP framed all of this under the banner of the “autonomous enterprise”. The pitch is that Joule agents can take real actions inside SAP processes rather than just answer questions about them. There is genuine substance under the marketing. Industry coverage from analysts like SAVIC Technologies on Joule Agentic AI tracks the rollout in detail.

This is the AI surface that SAP wants the market to associate with the SAP brand in 2026. Demos at Sapphire, keynote slides, partner marketing, all of it points at Joule.

## Where Joule is embedded today

Joule is bundled into the SAP enterprise stack. As of Q1 2026, that means:

- SAP S/4HANA Cloud (public and private editions). This is SAP’s flagship enterprise ERP and the deepest Joule integration.

- SAP SuccessFactors. The HR suite. Joule handles candidate screening, performance summary generation, learning recommendations.

- SAP Ariba. Sourcing and procurement. Joule agents draft RFx documents and analyze supplier responses.

- SAP Datasphere. The data platform. Joule writes queries and explains data lineage.

- SAP Intelligent Clinical Supply Management and a handful of newer industry clouds.

These are the products where Joule is a first-class citizen. A finance lead using S/4HANA Cloud in 2026 can ask Joule to draft an accrual entry and have it land in a workflow for review. That capability does not stop at the application boundary; the agents reach across SAP modules because they were built into the SAP Business Technology Platform from the start.

## Why Joule is not in SAP Business One

SAP Business One has a different origin story. SAP acquired it in 2002 from an Israeli company called TopManage. It was designed for small businesses, the segment SAP’s enterprise products had never served well. It runs on a separate code base, a separate database (HANA or Microsoft SQL Server), a separate partner ecosystem, and a separate roadmap. For more than twenty years that separation has worked fine. Business One is a respected SMB ERP. It has a healthy reseller channel. It ships steady updates.

But that separation also means SAP’s heavy AI investment in 2024 through 2026 has flowed into the enterprise products, not into Business One. The SAP Business Technology Platform, where Joule lives, is wired into S/4HANA. Building the same wiring into Business One would mean re-architecting a product whose customers chose it precisely because it is simpler and cheaper than S/4HANA. SAP has not made that investment, and there is no public roadmap suggesting it will. SAP’s Q2 2026 Business AI release highlights again cover the enterprise stack with no mention of Business One (checked August 2026).

The Business One product page in 2026 does not advertise Joule. Sapphire 2026 sessions on Business One focused on the SAP HANA migration story and partner add-ons, not on AI agents. The picture from SAP’s own materials is consistent: Joule is for the enterprise tier.

## What SAP Business One’s AI surface actually is in 2026

If you install Business One in 2026, here is what you get on the AI side:

- Nothing native from SAP. There is no built-in chat copilot, no agent runtime, no Joule sidebar.

- The Business One SDK. SAP has shipped an SDK for decades. Resellers and ISVs use it to build add-ons. Some of those add-ons now include AI features (document OCR, demand forecasting, chat over Business One data) sourced from OpenAI, Anthropic, or third-party Indian and European ISVs.

- Third-party integrations. Bridges to ChatGPT, Microsoft Copilot, and other external models exist as paid add-ons.

That is a real ecosystem and many of those add-ons are genuinely useful. But it is structurally different from Joule. With Joule on S/4HANA you get one vendor, one roadmap, one support contract, one set of agents that share context across finance and HR and procurement. With Business One you get a base ERP plus a shopping list of third-party AI add-ons, each with its own license, its own data flow, and its own support story.

## What this means if you are evaluating SAP Business One in 2026

If you are sizing up Business One, the SAP brand is doing some work in your head and that is fair. SAP has been around forever. The product is mature, the channel is large, the localization coverage is wide. None of that is in question.

What is worth being explicit about is the AI gap. When the reseller deck has a Joule slide, ask whether the deployment proposed for your business is S/4HANA or Business One. If it is Business One, the Joule slide is decorative. The actual AI footprint will be whatever third-party add-ons your reseller bundles in.

For SMB founders who have already decided they want a single AI-native stack rather than a base ERP plus add-on AI, Business One is probably not the right fit in 2026. The right fit is either S/4HANA Cloud (which means an enterprise budget and an enterprise implementation) or one of the smaller, newer products that are AI-native from line one.

## ERPClaw as one of those alternatives

ERPClaw is one option in that smaller-newer-AI-native bracket. Honest disclosure, I build it. The shape of the trade is different from Business One in a few ways that matter for an SMB:

- AI is the primary interface, not a sidebar. You type “invoice Acme $5,000 for April consulting, due in 30 days” into Telegram or a CLI, and the system parses, validates against the chart of accounts, posts the journal entries, and replies with the invoice number. No DocType screens to click through.

- Free forever, fully open source under GPL v3. Self-hosted on your own infrastructure, with no per-user fee and no vendor able to revoke access.

- 5-minute install. One command pulls the foundation and the modules you ask for. The system auto-detects your industry and pulls the right vertical pack.

- Cryptographic verification of every release. Each install resolves against a signed manifest, so you can prove the code you got is the code we shipped.

- Audit-grade accounting under the hood. Money is stored as exact decimals, never floats. Every posting passes a 12-step GL validation. Submitted entries are immutable; corrections are reversals, not edits.

The full comparison against the SAP enterprise products lives in 5 AI-native ERPs that earn the label and the AI-decorated vs AI-native software breakdown. The head-to-head against the SMB tier is at /compare/sap-business-one/.

Two related reads on competitor AI stories in 2026:

- Odoo’s Enterprise AI Pricing in 2026: What $31.10/user/month Actually Buys

- ERPNext AI in 2026: NextAI, ChatNext, ChangAI Compared

## The honest summary

SAP Joule is real, ambitious, and substantive. It is also not in SAP Business One, and there is no signal that it will be. If you are an SMB founder evaluating Business One in 2026, treat the Business One AI story on its own terms (third-party add-ons via the SDK ecosystem) and treat any Joule slide in a Business One deck as a brand association rather than a product feature. If you want AI-native architecture, the path runs either through the SAP enterprise tier or through one of the newer products that put AI at the center from the first commit.

Either choice is defensible. The wrong move is buying Business One on the assumption that the Joule demos from Sapphire are going to show up in your install.

Compare ERPClaw and SAP Business One side by side.

Tagssap-business-onesap-jouleai-erpcomparisonenterprise-erp
Share:XLinkedInHacker News

---

## Self-Hosted AI Accounting: Books on Hardware You Own
URL: https://www.erpclaw.ai/blog/self-hosted-ai-accounting-2026/

# Self-Hosted AI Accounting: Books on Hardware You Own
What self-hosted AI accounting actually means, why data sovereignty matters for your books, what it looks like day to day, and when self-hosting is the wrong call.

Most tools that call themselves “AI accounting” keep your books on the vendor’s servers. You type into a web app, the data lives in their cloud, and an AI feature sits on top of it. Self-hosted AI accounting flips that: your books live on a machine you control, and the AI does the bookkeeping work right there against your own data. Nothing about your finances leaves your hardware unless you choose to send it.

ERPClaw is AI-native, which means the AI is the primary user of the system, not a chat box bolted onto an old database. If you want the plain-English basics of how AI accounting works at all, start with the AI accounting explainer. This post is narrower and more practical: what running it on your own hardware really looks like, who it suits, and who should skip it.

## What self-hosted AI accounting actually means

Two ideas get stacked together in the phrase “self-hosted AI accounting,” and it helps to pull them apart.

Self-hosted is about location and control. Your accounting data, the customer records, the invoices, the bank and payout history, the payroll figures, sits on hardware you own or rent directly. There is no vendor account holding the master copy. You are not logging into someone else’s website to see your own numbers. The software runs where you put it, and the data file is yours to copy, back up, move, or delete.

AI-native is about who does the work. In an AI-native system, the AI drafts the invoice, proposes the reconciliation, posts the journal entry, and runs the report. You describe what you want in ordinary language and the agent carries it out against your real ledger. That is different from the AI-decorated approach most cloud accounting tools take, where a chat feature summarizes a dashboard or suggests a category but a human still drives every meaningful posting. The architecture argument behind that distinction lives at AI-native ERP and in AI-decorated vs AI-native software; I will not re-fight it here.

Put the two together and self-hosted AI accounting means this: the AI does the accounting work, and it does that work on books that physically live with you.

## Why data sovereignty matters for your books

Your books are some of the most sensitive records your business holds. Customer names and contact detail, payment history, payroll, bank account numbers, vendor contracts. On a cloud accounting product, all of that sits on the vendor’s servers, under the vendor’s terms of service, in a tenancy you share with thousands of other companies.

That arrangement works fine right up until it doesn’t. A price increase is the vendor’s decision, not yours. An acquisition can change who owns the company holding your financial history. An outage means you cannot see your own numbers until someone else fixes their system. A change to data-handling terms can happen with an email you skim and a checkbox you click. None of those are hypothetical; they are the normal life cycle of a software company you do not control.

Private AI accounting on your own infrastructure removes that whole category of exposure. Your data stays where you put it. The AI reads and writes against a local copy. Nothing phones home. When you want to send a figure to your accountant or your bank, you send it deliberately, and only that figure.

This matters more in some contexts than others. A solo founder may simply prefer not to scatter financial detail across cloud accounts they will forget to cancel. A small CPA practice carries a client-confidentiality duty and may have clients who expect their books to stay off shared vendor servers entirely. A business with customer-data obligations wants a clear answer to “where does this information live,” and “in a vendor’s multi-tenant cloud” is a harder answer to give a worried client than “on a machine in our office.” Self-hosting gives a simple, honest answer: here, on this machine. I am not claiming a specific compliance certification for you; I am pointing out that “your data is on your hardware” is a much easier sentence to defend than “our data is somewhere in a vendor’s multi-tenant cloud.” If the larger rent-versus-own economics interest you, the open source vs SaaS ERP question covers that ground.

## What self-hosted AI accounting looks like day to day

This is the part nobody seems to write about, so here is the operator’s-eye view.

You install it once. The install runs in a few minutes on a machine you already have. There is no account to create, no card to enter, no trial clock.

You describe your business in plain English. You tell the agent what kind of business you run, and it sets up the accounts and structure to match. You do not configure a chart of accounts by hand or hunt through settings menus.

The AI does the daily work. “Create an invoice for Wayne Enterprises, twelve thousand five hundred dollars, net thirty.” “Reconcile last week’s Stripe payout.” “Show me the profit and loss for May.” You ask, the agent drafts or runs it against your local books, and you confirm. The bookkeeping happens against your real ledger, in seconds, without you touching a spreadsheet.

You back up by copying a file. Your entire set of books is a file on your machine. Backing up is copying that file somewhere safe. Restoring is copying it back. There is no export wizard and no “request your data” support ticket.

You run reports by asking. Trial balance, aging, profit and loss by segment, what you actually made last month. You ask in ordinary language and the agent runs the query against your real ledger and returns a clean answer. There is no report builder to learn and no menu to hunt through, because the question is the interface.

Month-end is a conversation, not a project. When you close the books, you ask the agent to reconcile the accounts that need reconciling and surface anything that does not line up. It drafts the routine entries and flags the handful of items that need your judgment. The judgment calls stay with you, where they belong; the volume of matching and posting gets done while you watch.

One honest note on the experience today: the primary way you work with the system is through an AI agent in a terminal. That is a real interface and it is fast, but it is text, not a polished point-and-click web app. The web dashboard exists but is developer-only for now. If a clickable screen for non-technical staff is a hard requirement this quarter, factor that in. And the price stays the same regardless of how much you use it: $0 forever, self-hosted on your own infrastructure.

## Self-hosted on a laptop, a server, or PostgreSQL

“Self-hosted” does not mean one fixed setup. It scales with you.

On a laptop. For a solo founder or a one-person shop, the whole thing runs comfortably on the laptop you already work on. Your books, the AI, and your day-to-day all live in one place.

On a small office server. For a small team, you can put it on a shared machine in the office or a server you rent, so more than one person works against the same books.

On PostgreSQL. For heavier use, more data, or several people writing at once, the same software runs on PostgreSQL instead of the default local database. This is worth underlining because people assume self-hosted means a toy database that falls over. It does not. PostgreSQL is a first-class option, not a fallback, and the same code runs on either backend. Moving to it is a configuration change, not a rewrite or a migration to a different product.

For the deeper, more technical case on how the data layer is built and when each backend makes sense, the SQLite or PostgreSQL database choice post is written for the engineer on your team. For everyone else, the takeaway is simpler: start where you are, grow without switching products.

## When self-hosting is the wrong choice for you

I would rather you make a good decision than make our decision, so here is the honest part. Self-hosted AI accounting is genuinely the wrong fit in several real situations.

Nobody can run a backup. If there is no one in the business who is comfortable copying a file to a safe place on a schedule, the data-ownership benefit turns into a data-loss risk. Paying for a managed product that does it for you is the right call, and you can have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

Your procurement rules require a contracted SaaS vendor. Some companies, and many that sell into the enterprise, have a mandate: the accounting system must be a named SaaS vendor with a signed service-level agreement and, sometimes, an AICPA-recognized status. If that is your reality, a cloud product is the right answer and self-hosting will not pass review.

You need a polished web interface for non-technical staff today. If several people who are not comfortable with a terminal need to click around in the books this quarter, the agent-first experience will frustrate them. A traditional web-app accounting product fits better right now.

You want a vendor to call at 2am. Self-hosting means you own the uptime. If you want a support line that owns the system being available, a managed vendor is the trade you want.

In all of those cases, a managed cloud product is the correct call, and you should make it without guilt. Unless your procurement rules name the vendor for you, that managed product can be ours: ERPClaw Cloud Managed, onboarding by request through /pricing/. Self-hosting earns its keep when you value ownership and privacy, have a basic comfort with running your own setup, and want the AI to do the volume of the work.

## Frequently asked questions

What does self-hosted AI accounting mean in plain terms?
It means two things at once. Your books live on a machine you control instead of a vendor’s cloud, and an AI does the bookkeeping work, drafting invoices, reconciling payments, running reports, against those local books. You describe what you want in plain English and the agent carries it out on your own data.

Is my financial data private if I self-host?
Yes. Your data sits on your hardware, and the AI reads and writes against that local copy. Nothing leaves your machine unless you choose to send it. There is no shared multi-tenant cloud holding the master copy of your books, which removes the data-segregation and vendor-terms exposure that comes with cloud accounting. For the broader view on where AI should and should not act on your books, see where AI belongs in accounting.

Do I need to be technical to self-host my accounting?
You need a basic comfort with running software you installed yourself. The daily work is done by talking to an AI agent in plain language, so the accounting itself does not require technical skill. The honest caveat is that the main interface today is a terminal agent rather than a polished web app, and you should be willing to copy a backup file on a schedule. If neither of those is a problem, you are technical enough.

Can I move from a laptop to a server later without redoing everything?
Yes. You can start on a laptop and later move to a shared server, or move to PostgreSQL when you outgrow the default local database. The same software runs on either backend, so switching is a configuration change rather than a rewrite or a migration to a different product. You do not redo your books to grow.

What happens to my books if the project stops being maintained?
You keep everything. ERPClaw is open source, so the code is yours to keep running, and the data file is already on your machine. If the project ever stopped, you would still hold both the working software and your complete books, and the community could continue the code. That is the opposite of a cloud product, where the end of the vendor can mean the end of your access.

When should I use cloud accounting instead of self-hosting?
Use cloud accounting when nobody in the business can reliably run a backup, when your procurement rules require a contracted SaaS vendor with a service-level agreement, when several non-technical staff need a polished web interface today, or when you want a vendor to own uptime and answer a support line at 2am. In those cases a managed cloud product is the right call, and self-hosting is not. The managed product can still be ERPClaw: Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

Does the AI send my data anywhere to work?
The agent reads and writes against your local books on your own machine. The work of drafting an invoice, reconciling a payout, or running a report happens against that local data. The only thing that leaves your hardware is what you deliberately send out, such as a figure you forward to your accountant or an invoice you email to a customer. Self-hosting is what makes that line clear: by default, your financial detail stays put.

## The short version

Self-hosted AI accounting is for people who want the AI to do the bookkeeping work and want the books on hardware they control. You get privacy and ownership, you grow from a laptop to a server to PostgreSQL without switching products, and it costs $0 forever, self-hosted on your own infrastructure. It is not for everyone, and the section above is honest about when a cloud product wins instead.

If that sounds like your business, talk to a co-founder in a quick demo, or read the AI accounting explainer for how the whole thing works. Your other open-source options are scored honestly in the open source AI accounting roundup, and the wider category context is in the state of AI in accounting.

Tagsself-hostedai-accountingdata-sovereigntyopen-source
Share:XLinkedInHacker News

---

## Shopify Payout Reconciliation: Why It Never Matches
URL: https://www.erpclaw.ai/blog/shopify-payout-reconciliation-explained/

# Shopify Payout Reconciliation: Why It Never Matches
Your Shopify deposit doesn't match your sales total. Here's why Shopify payout reconciliation is a mess, and the simple pattern that fixes it.

It’s Saturday morning. You open your bank app and see a deposit from Shopify for $4,318.62. Then you open your Shopify dashboard, look at last week’s sales, and see $5,247.00. The two numbers don’t match. They’re never going to match. You sigh and open a spreadsheet.

If that scene feels familiar, this post is for you. Shopify payout reconciliation is the most confusing piece of running a Shopify store, and almost nobody explains it in plain English. We’ll fix that: why your Shopify deposits never match your sales totals, the simple accounting pattern that makes the math work, and how to stop doing this by hand forever.

## The Saturday morning spreadsheet problem

Here’s what most Shopify owners actually do on the weekend:

- Pull last week’s payout amount from the Shopify dashboard.

- Pull last week’s order list. Add up the totals.

- Subtract Shopify’s transaction fees. Guess at the percentage.

- Subtract refunds. Wait, did that refund hit this payout or the next one?

- Numbers still don’t match. Open a beer.

This isn’t a moral failing. A single Shopify deposit is the net result of dozens (sometimes hundreds) of separate financial events that all happened at different times. Until you understand how Shopify actually moves your money, no amount of staring at a spreadsheet will make it click.

## Why Shopify payouts confuse merchants (it’s not you, it’s the math)

When a customer buys a $100 t-shirt from your store, several things happen, and only one of them is “you got $100.”

Shopify Payments takes its cut. For a US store on the basic plan, that’s roughly 2.9% plus 30 cents per card transaction. So your $100 sale becomes about $96.81 owed to you, with $3.19 going to Shopify as a transaction fee.

The money sits in Shopify’s account for a few days. Shopify holds funds before depositing them. The standard payout schedule is daily with a 2-3 business day delay. So a sale on Monday might not land in your bank until Thursday or Friday.

Refunds, exchanges, and chargebacks get netted out. If a customer returns last week’s purchase today, Shopify subtracts the refund from your next payout. So today’s deposit is funded by sales from three days ago, minus refunds processed today, minus any disputes pending.

Some funds get held in reserve. If you sell anything risky (preorders, high refund rates, a sudden volume spike), Shopify Payments may hold a percentage in reserve. That reserve doesn’t show up in your payout but it does count as money owed to you.

Multiple payouts can settle in a single bank deposit. Or one payout can be split. Or a payout can be in a different currency if you sell internationally.

So the deposit on your bank statement isn’t “last week’s sales minus fees.” It’s a tangled net of money from various days, with various deductions, all summed into one wire. Gross sales, net sales, payout amount, and bank deposit are four different numbers. If you try to reconcile by matching them directly, you’ll fail. You need a different pattern.

## The Shopify clearing account: the only pattern that actually works

The trick that real accountants use, and the trick built into ERPClaw’s Shopify payout reconciliation engine, is called a clearing account. The name sounds intimidating but the idea is simple.

Think of a clearing account as a holding bucket between Shopify and your bank account. Every time you make a sale, the money gets credited to this bucket on paper. Every time Shopify takes a fee or processes a refund, that comes out of the bucket. And every time Shopify deposits money in your real bank account, that gets pulled out of the bucket too.

If everything is recorded correctly, the bucket should always end up empty. That’s the whole point. If the bucket has money sitting in it, that’s money Shopify still owes you. If the bucket goes negative, something was double-counted. The bucket is your truth-teller.

Here’s the math on a single $100 sale, written out for someone who doesn’t speak accountant.

### Step 1: The sale happens

Customer buys a $100 t-shirt. Shopify will keep $3 in fees and owe you $97.

In your books:

- Your sales for the day went up by $100.

- Your “Shopify owes me” bucket (the clearing account) went up by $97.

- Your “Shopify fees” expense went up by $3.

The bucket now holds $97.

### Step 2: Shopify holds the money for a couple of days

Nothing changes in your books. The bucket still holds $97. Your bank account hasn’t moved yet. Normal.

### Step 3: Shopify deposits the payout

Three days later, $97 lands in your real bank account.

In your books:

- Your bank account went up by $97.

- Your “Shopify owes me” bucket went down by $97.

The bucket is now empty. It’s fully reconciled. The $100 sale, the $3 fee, the $97 deposit, all of it lines up. You can prove every dollar.

That’s it. That’s the entire pattern. Now imagine the same logic running on every order, every refund, every chargeback, every payout, every day, automatically. The bucket should always sit at zero (or close to it, with just the in-flight money sitting there for 2-3 days).

If you want the formal version with debits and credits, the Shopify integration architecture page walks through the journal entries. For this post, the bucket metaphor is enough.

## Why this matters for Shopify payments accounting

Most Shopify owners (and a surprising number of bookkeepers) skip the clearing account entirely. They take the bank deposit, call it “Shopify income,” and move on. This works fine until tax season, when your accountant asks three questions:

- What was your gross revenue? You don’t know. You only recorded the net deposits.

- How much did you pay in Shopify fees? You don’t know. They were never broken out.

- What about that refund last March? You don’t know. It just made the deposit smaller.

The IRS wants gross revenue on your tax return, not net deposits. Your accountant wants to deduct Shopify fees as a business expense. Your investors want to see your real margins. None of that is possible if you’re booking the bank deposit as your sales number.

The clearing account fixes all of this. Gross revenue is recorded the moment the sale happens. Fees are tracked separately. Refunds reverse cleanly. The bank deposit is just the deposit, not a stand-in for sales. If you’re trying to reconcile Shopify payouts to QuickBooks, this is also the only pattern that scales: without a clearing account you end up with hundreds of generic “deposit from Shopify” lines that no human can audit.

## The three-layer reconciliation, in plain English

Once you have the clearing account, real Shopify payment accounting becomes a three-layer check that runs every day. ERPClaw runs all three automatically; if you’re doing it by hand, here’s what to verify.

### Layer 1: Did every order make it into the books?

For every order Shopify recorded, there should be one matching entry in your accounting system. Same order ID, same total, same date. If Shopify says you had 47 orders yesterday, your books should show 47 entries from yesterday. If one is missing, the sync failed for that order and you need to find out why.

This is the easy layer. It catches sync gaps.

### Layer 2: Did every payout get matched?

For every Shopify payout, there should be a corresponding bank deposit. Same date, same amount. If Shopify says it paid you $4,318.62 on Tuesday and your bank shows $4,318.62 from Shopify on Wednesday, you’re good. If the amounts don’t match, the most common reason is currency conversion or a wire fee from your bank.

This layer catches deposit mismatches and Shopify payouts not matching your bank.

### Layer 3: Does the clearing account zero out?

Add up everything that went into the bucket (orders) and everything that came out of the bucket (fees, refunds, payouts, reserves). The number should be zero, or close to it (the only money that should be sitting in the bucket is sales from the last 2-3 days that haven’t been paid out yet).

If it doesn’t zero out, something is wrong. A fee got missed. A refund got booked twice. A payout got recorded but the underlying orders didn’t sync. Whatever it is, the clearing account is the canary. It tells you there’s a problem before your accountant does.

This three-layer check is the foundation of every serious Shopify payout reconciliation system, including the one built into ERPClaw.

## When reconciliation breaks: the four most common causes

Every Shopify owner eventually hits a “Shopify deposit not matching” moment. After supporting hundreds of stores, almost every case fits one of these four patterns.

### 1. Currency conversion

You sell in USD, but a customer paid in EUR. Shopify converted the EUR to USD using its rate, your bank converted again on the way in using its rate, and the two rates disagree by a tenth of a cent. Net result: your payout is $4,318.62 and your bank deposit is $4,316.91. The $1.71 difference is currency conversion, not a missing transaction.

The fix is to record the difference as a foreign exchange gain or loss, not to chase it.

### 2. Refund timing

A customer returned a $50 item on Monday. Shopify processed the refund Tuesday morning. The refund came out of Tuesday’s payout, but you booked the original sale on Monday. So Tuesday’s payout is “$50 lower than expected” if you’re comparing to Tuesday’s orders, but “$50 higher than expected” if you’re comparing to Monday’s orders.

The clearing account makes this a non-issue: the refund just reverses the original sale’s contribution to the bucket. Without a clearing account, refund timing will haunt you forever.

### 3. Disputes and chargebacks

A customer disputes a $200 charge. Shopify pulls $200 plus a $15 dispute fee from your next payout and holds it as a reserve until the dispute resolves. Your payout is $215 lower than your sales would predict. If you win, the $215 comes back weeks later. If you lose, it’s gone. Either way, you need a “Dispute Reserve” line on your books, separate from sales and fees. ERPClaw handles this automatically.

### 4. App fees, capital repayment, and Shopify subscription

Shopify Capital repayments come straight out of payouts. Shopify subscription fees do too. Apps you’ve installed sometimes bill through Shopify and get netted out. These all reduce your payout below “sales minus transaction fees” and each needs its own category to be deductible at tax time. Most middleware tools lump them together as “other deductions.” Track them separately.

## How ERPClaw automates Shopify payout reconciliation in one command

We built ERPClaw because doing all of the above by hand is unsustainable past about 50 orders a month. Here’s what the automated version looks like.

You install ERPClaw on your own machine (laptop, Mac mini, cheap VPS, your call). You pair the ERPClaw Shopify integration to your local install with a six-character code. The integration is shipped and installable today; the docs walk through both the pairing flow and the self-hosted Custom App token flow. Total setup time is under ten minutes; the install walkthrough covers every step.

After that, every Shopify event flows automatically:

- Orders post to a Shopify Clearing account, with revenue and inventory tracked separately.

- Fees post to an expense account in real time.

- Refunds reverse the original entries cleanly.

- Payouts move money out of the clearing account into your bank account.

- Disputes create reserve entries and resolve themselves when the dispute closes.

- Gift cards post as deferred revenue (a liability) and convert to revenue only when redeemed.

Then you run one command (or just ask the assistant in plain English) and ERPClaw shows you:

`Shopify Clearing balance: $1,247.18 (in-flight, last 2 days)
Last 30 days reconciliation:
Orders synced: 1,247 / 1,247 OK
Payouts matched: 34 / 34 OK
Clearing nets to: $0.00 OK`
Three green lights. That’s the whole reconciliation. If anything is off, ERPClaw tells you which order, payout, or fee is the problem and gives you the exact entry to look at.

The whole thing is open-source, AI-native, runs on your machine, and costs $0. Not a free trial, just $0 forever. The other tools in this category (A2X, Bookkeep, Synder, Webgility) are all AI-decorated middleware built before AI was practical; ERPClaw is the only AI-native option, which means you ask the assistant questions like “show me which payouts didn’t fully reconcile last month” in plain English. If you’ve been paying $29 to $229 a month (A2X pricing, as of June 2026) for middleware that does roughly the same job (and only outputs to QuickBooks or Xero), the A2X comparison walks through what you get for free.

## Comparison: doing it by hand vs. middleware vs. ERPClaw

| Approach
| Setup time
| Monthly cost
| Catches errors
| Tax-ready books

| Spreadsheet by hand
| 0 hours
| $0
| No, you find them in April
| Maybe

| A2X / Bookkeep / Synder
| 1 hour
| $29-229
| Summary level only
| Yes (in QBO/Xero)

| ERPClaw
| 10 min
| $0
| Transaction level + clearing
| Yes (in ERPClaw)

The middleware tools are fine if you already live in QuickBooks. ERPClaw is the right pick if you want the full picture in one place, want to own your data, and don’t want to keep paying for something the open-source community can ship for free. See the pricing page for the full breakdown.

## FAQ

### Why doesn’t my Shopify deposit match my sales total?

Because they’re measuring different things. Your sales total is gross revenue. Your deposit is gross revenue minus Shopify fees, minus refunds processed during the payout window, minus any reserves, minus capital repayments and subscription fees. They will never match without a clearing account to reconcile them.

### What is a Shopify clearing account?

A clearing account is a temporary holding bucket on your books that tracks money Shopify owes you. Every sale increases the bucket. Every fee, refund, and payout decreases the bucket. When the bucket is empty, you’re fully reconciled.

### How do I reconcile Shopify payouts to QuickBooks?

The standard approach is to use a clearing account in QuickBooks called “Shopify Clearing” (or “Undeposited Funds, Shopify”). Post each order’s net amount to the clearing account, post each payout as a transfer from clearing to your bank account, and verify that the clearing balance returns to zero each cycle. Tools like A2X automate this by posting summary journal entries; ERPClaw automates it at the transaction level.

### How long do Shopify payouts take to hit the bank?

Standard US payouts are 2-3 business days from the date of the sale. New stores or stores in higher-risk categories may have longer hold periods. International payouts and currency conversion add time too. See the official Shopify Payments documentation for the full schedule.

### What if my Shopify Clearing account doesn’t zero out?

Either money is in flight (sales from the last 2-3 days haven’t been paid out yet, which is normal) or there’s an error. Common errors: a fee that wasn’t recorded, a refund that was double-counted, a payout that was booked but its underlying orders didn’t sync, or a currency conversion gap. Walk through the three-layer check above to find which one it is. The Shopify troubleshooting guide covers the most common cases step by step.

### Do I need a clearing account if I just use QuickBooks?

Yes. QuickBooks doesn’t know how Shopify’s payout structure works on its own. Without a clearing account, you’ll end up with bank deposits that don’t tie to anything, fees that aren’t deductible, and a tax filing your accountant will charge you extra to clean up.

## Stop reconciling Shopify by hand

You didn’t start a Shopify store so you could spend Saturday mornings in a spreadsheet. The clearing account pattern is the single biggest fix in ecommerce accounting, and there’s no reason to keep doing it manually in 2026.

ERPClaw runs the full three-layer reconciliation automatically, on your own hardware, for free. Install it, pair it to your store with one command, and watch your Shopify Clearing account zero out for the first time. It’s a strangely satisfying feeling.

Get started: install the ERPClaw Shopify integration or read the 10-minute install walkthrough. Comparing to a paid tool? The A2X comparison shows where the dollars go, and if you plan to stay on middleware, A2X vs Synder settles which connector fits.

---

## SQLite or PostgreSQL: Why ERPClaw Defaults to SQLite
URL: https://www.erpclaw.ai/blog/sqlite-or-postgresql-database-choice/

# SQLite or PostgreSQL: Why ERPClaw Defaults to SQLite
ERPClaw is database agnostic via PyPika. SQLite is the default for self-hosted installs, PostgreSQL is fully supported for enterprise workloads. Here is why.

ERPClaw works with SQLite or PostgreSQL. PyPika abstracts the query layer so the same code runs on either backend. Most installs use SQLite by default. Here is why.

This post is for the CTO or engineering manager who has shipped enough PostgreSQL-backed enterprise software to know why people pick it, and is now trying to evaluate whether a self-hosted ERP that defaults to SQLite is serious or naive. The short answer: SQLite is the default because it fits the workload of a 40 to 500 employee self-hosted install. PostgreSQL is a first-class option for the workloads that need it. The architecture supports both because the abstraction was built that way from day one.

## How database agnosticism works in ERPClaw

Every database query in ERPClaw goes through a thin abstraction built on PyPika, a Python query builder that emits SQL for multiple dialects. PyPika is dialect-aware, so the same Python code that constructs a query against SQLite in a self-hosted install constructs the equivalent query against PostgreSQL in an enterprise deployment. There is no second codebase, no dialect-specific branching scattered through the action handlers, and no rewrite when a customer crosses the threshold where PostgreSQL is the better fit.

Switching engines is a configuration change, not a refactor. The connection layer reads the target backend from environment configuration, hands back a connection object that honors the same interface, and the rest of the application is unaware of which engine is underneath. Schema DDL ports with mechanical translation. The 12-step GL validation, the Decimal-based money handling, the immutable GL semantics, the constitutional rules engine all run unchanged on either backend.

The rest of this post argues SQLite is the right default. The point of starting here is that you do not have to take that on faith: if the workload outgrows the default, the migration path is mechanical, not architectural.

## The reflex to pick PostgreSQL

Every back-end engineer who has built a multi-user system in the last fifteen years has the same default. Pick PostgreSQL. It is the safe answer, the answer that survives an enterprise architecture review, the answer the RFP expects.

That reflex is correct for a wide range of workloads. PostgreSQL is an exceptional database. It scales horizontally with read replicas, it has the best feature set of any open-source RDBMS, and the operational story around it (backups, failover, monitoring) is mature. For a SaaS product with thousands of concurrent writers across hundreds of tenants, PostgreSQL is the right answer, and ERPClaw supports it for exactly those deployments.

The reflex becomes worth questioning when the workload no longer matches the reasoning. A self-hosted ERP for a 40 to 500 employee company has roughly these characteristics: one process writing at a time, bursts of dozens of reads in parallel, total daily write volume well under 100,000 rows, total database size under 10 GB for the first several years, and a single operator who is not a DBA. Most of the assumptions that make PostgreSQL the obvious answer do not apply, which is why SQLite is the default.

Three forces keep PostgreSQL as the assumed default even when SQLite is the better fit:

- Perception of “real” databases. Many developers were taught that SQLite is for mobile apps and CI fixtures. SQLite the project has never tried very hard to correct this perception, and the result is a generation of engineers who believe a single file database cannot run production.

- Concurrency myths. The claim that SQLite “cannot handle concurrent writes” is technically true and practically misleading. It serializes writes within a single process. For a workload where writes are already being serialized by the application, this is a feature, not a limitation.

- Vendor RFP requirements. Enterprise procurement forms ask for PostgreSQL or Oracle by name. When the procurement reflex matters, ERPClaw runs on PostgreSQL and the form gets the answer it expects.

The first two are outdated assumptions. The third is a real constraint, and the dual-backend architecture is how ERPClaw addresses it directly.

## What SQLite actually does well in 2026

The SQLite documentation has a famous page called Appropriate Uses For SQLite that has not changed much in years, because the answer has not changed much. It is worth re-reading with fresh eyes if your last encounter with SQLite was a Django tutorial.

WAL mode and concurrent reads. SQLite in Write-Ahead Logging mode gives you concurrent readers and a single writer with no reader blocking. A reporting query running against the trial balance does not block a sales invoice submission, and vice versa. This is the configuration ERPClaw runs in production:

`conn.execute("PRAGMA journal_mode = WAL")
conn.execute("PRAGMA synchronous = NORMAL")
conn.execute("PRAGMA foreign_keys = ON")
conn.execute("PRAGMA busy_timeout = 5000")`
Four pragmas. That is the entire production tuning surface for a database that handles double-entry bookkeeping, inventory, payroll, and the full operations surface of a real ERP.

Foreign key enforcement. SQLite supports declarative foreign keys with cascade rules, and once you set `PRAGMA foreign_keys = ON`, the enforcement is identical to what you would expect from PostgreSQL. The reason this pragma is off by default is historical compatibility, not a missing feature.

ACID guarantees that match most managed cloud databases. SQLite is fully ACID compliant. Its durability story (when configured correctly) is stronger than several managed cloud databases that quietly default to weaker consistency for performance. A `COMMIT` in SQLite means the bytes are on disk. A `COMMIT` in some hosted Postgres-flavored services means the bytes are on the way to disk on a replica that may not have acknowledged yet.

No network round trip. Every query in SQLite is an in-process function call. There is no socket, no protocol parsing, no connection pool tuning, no PgBouncer to deploy. For a workload where the database and the application live on the same machine, this is a 100x latency improvement on small queries that happen thousands of times per workflow.

Operational simplicity. A SQLite database is a file. Backup is `cp data.sqlite backup.sqlite` while the database is in WAL mode. Restore is the inverse. There is no `pg_dump`, no `wal_archive`, no replication slot to manage. For a self-hosted ERP whose operator is the founder of a 50 person company, this matters more than any feature comparison.

The result is a database that, for the workload most ERPClaw installs have, is faster, simpler, more durable, and cheaper to operate than a managed alternative. The reason most engineers do not believe this is that they are reasoning about a different workload, the one PostgreSQL is the right answer for.

## Where SQLite falls short (and PostgreSQL takes over)

This is the section the rest of the internet skips, so let me be specific. Each of the following is a real limit of SQLite, and each is the reason the PostgreSQL backend exists in the same codebase.

Write concurrency under contention. SQLite serializes writes. If two processes try to begin a write transaction at the same time, the second one waits up to `busy_timeout` milliseconds and then fails with `SQLITE_BUSY`. For a workload where many processes are writing concurrently (a public web app with thousands of users hitting the same tables), this is a real limitation, and the fixes (sharding, queueing, retries) get awkward fast. PostgreSQL is the right answer.

Massive parallel write throughput. SQLite tops out at low thousands of writes per second on commodity hardware, single-process. PostgreSQL with proper tuning can sustain tens of thousands of writes per second across many connections. For workloads that cross that threshold, switch the backend.

Read replicas. SQLite has no native replication story. You can use Litestream for streaming backups to S3, and projects like LiteFS provide replicated SQLite, but the ergonomics are nowhere near what PostgreSQL gives you for free with streaming replication. If your compliance regime or DR posture requires synchronous standbys, run PostgreSQL.

Network access. SQLite is an embedded library. If you need a database that multiple application servers connect to over the network, SQLite is not the answer without an additional layer. PostgreSQL is built for that shape.

Server-side functions and extensions. PostgreSQL has stored procedures, triggers, JSON operators, full-text search, GIS, and a thousand extensions. SQLite has a useful but smaller set. For analytical workloads or anything that wants the database to do heavy lifting, PostgreSQL wins.

If your workload looks like any of those, configure ERPClaw with the PostgreSQL backend. The next two sections explain why most self-hosted ERPClaw workloads do not look like any of those, which is why SQLite is the default rather than the only option.

## How ERPClaw absorbs the SQLite tradeoffs

The default deployment was designed around the SQLite write model, not in spite of it. The architectural choices that make this work are visible in the developers documentation and reproducible in the open-source codebase.

Single-process write serialization. Every ERPClaw write goes through a single Python process talking to a single SQLite file. The AI agent submits an action, the action runs as one transaction, and the next action waits for the first to commit. Submit operations in an ERP are inherently transactional (a sales invoice writes to `sales_invoice`, `gl_entry`, `stock_ledger_entry`, and `journal_entry` atomically), and serializing them is the correct semantics. The application would have to serialize them anyway to satisfy the 12-step GL validation, so letting SQLite do it for free is a win.

5000 ms busy_timeout. The `PRAGMA busy_timeout = 5000` setting tells SQLite to wait up to five seconds for a lock before failing. For a workload where the longest write transaction is on the order of tens of milliseconds, that headroom means `SQLITE_BUSY` is essentially never observed in production. When it does happen, it is a signal of a bug (a transaction held open across an external API call, for example).

Concurrent reads under WAL. Reporting queries, dashboard refreshes, and the AI agent’s read-side scans all run concurrently with writes thanks to WAL mode. A trial balance query that takes 200 ms does not block the next sales invoice submission. This is the property that makes the single-writer model acceptable for an interactive application.

The PostgreSQL backend, ready when needed. Because PyPika and the connection layer abstract the dialect, the move to PostgreSQL is a configuration change, not a rewrite. The full migration playbook will get its own engineering post, and the features overview documents both stores.

## The operational benefits of the SQLite default

I want to give you the operational picture, because this is where the SQLite case stops being theoretical and starts mattering to whoever is going to run the system at 2 a.m. on a Saturday.

| Operation
| SQLite (ERPClaw default)
| PostgreSQL (typical self-hosted)

| Cold start
| < 50 ms
| 5 to 30 seconds

| Backup
| `cp data.sqlite backup.sqlite`
| `pg_dump` with role and password setup

| Restore
| `cp backup.sqlite data.sqlite`
| `pg_restore` with target DB created first

| Disk footprint (empty)
| 1 MB
| 40 to 60 MB

| RAM at idle
| < 10 MB
| 100 to 300 MB depending on config

| Network exposure
| none (file on disk)
| port 5432, needs firewall and auth

| Connection pooling
| not needed
| usually requires PgBouncer

| Schema migration tooling
| direct DDL
| Alembic or similar

| Operator skill required
| basic Unix
| DBA familiarity helpful

Some of these are unfair to PostgreSQL on a properly managed cluster, and some are unfair to SQLite (the lack of network exposure is a feature, not a deficiency). The point is the cumulative shape: the SQLite default eliminates entire categories of operational concern. There is no DBA. There is no replication lag. There is no `pg_hba.conf` to tune. The dev / prod parity is total, because dev and prod use the same single-file database and the same four pragmas. When you switch to the PostgreSQL backend, you take on the operational surface in exchange for the capabilities that come with it.

## Self-hosted ERP runs on a laptop

An ERPClaw instance with a year of transactions, twenty users, three integrations, and a full GL fits in a few hundred megabytes on a developer’s laptop. You can copy it to a USB stick. You can run it offline. You can run a full integration test against a fresh fixture in under thirty seconds.

That is not because ERPClaw is small. The schema spans the full operations surface from books to inventory to payroll, the action surface is deep, and the test suite covers thousands of assertions on every commit. It is because SQLite makes the runtime overhead of a database engine almost zero, and because a single file is a fundamentally different operational primitive than a database server. SQLite is what makes `clawhub install erpclaw` work as a one-line setup. PostgreSQL is the right backend for a public SaaS with 50,000 tenants, which is why it lives behind a backend flag for the deployments that want it.

## When PostgreSQL is the right backend

The honest version of any architecture post needs a section that says “default to the other option, in these cases.” For ERPClaw, the cases are clear, and PostgreSQL is a first-class choice for each of them.

Enterprise multi-tenant deployments with shared database. If you are running ERPClaw as a hosted product where many tenants share one database and you need row-level security, sophisticated indexing across tenant boundaries, and online schema changes that do not block writers for long stretches, PostgreSQL is the right backend. SQLite per-tenant works for some shapes (one file per tenant) but the ergonomics get awkward past a few hundred tenants.

More than 1 million write events per day on a single instance. SQLite can absorb a lot more than people assume, but if your sustained write rate puts you in the millions per day on a single database, PostgreSQL with proper tuning, connection pooling, and partitioning is the right answer. Configure the backend at install time and the rest of the application does not change.

Hard read-replica requirements. If your workload requires geographically distributed read replicas with bounded lag, or if your compliance regime requires synchronous replication to a standby, PostgreSQL has spent twenty years getting this right. ERPClaw rides on top of that maturity through the same abstraction layer.

Regulatory backup and replication policies. If your compliance posture mandates point-in-time recovery with a specific RPO, WAL archiving to a managed store, or audited replication to a regulated jurisdiction, PostgreSQL’s tooling is the path of least friction. Run the PostgreSQL backend and inherit the ecosystem.

Heavy analytical queries against the OLTP store. SQLite can do joins and aggregations at respectable speed, but it does not have a query planner that handles 200 line analytical queries the way PostgreSQL does. If your reports are doing serious analytics over millions of rows, run PostgreSQL or, better, replicate to a separate analytical store like DuckDB or ClickHouse.

Many writer processes. If your application architecture has multiple processes writing to the same database concurrently and you cannot serialize them through a single front-end, PostgreSQL is the right backend. The contention model is what you want.

ERPClaw’s roadmap includes a managed cloud edition where the workload genuinely will cross some of these thresholds. That edition runs on PostgreSQL via the abstraction layer described earlier. The single-tenant self-hosted edition stays on SQLite as the default because that is what the workload calls for. One application, two stores, no compromise.

## FAQ

Is SQLite really safe for production financial data?

Yes, when configured correctly. SQLite is one of the most thoroughly tested pieces of software on Earth. The test suite runs to over 100x the size of the source code. With WAL mode, `synchronous = NORMAL`, and foreign keys enabled, the durability guarantees match what you would expect from any serious RDBMS. The constitutional rules engine in ERPClaw adds another layer: every GL posting is verified against twelve invariants before commit, and the trial balance is checked across the entire database after every test run. The same invariants run against the PostgreSQL backend, because they sit above the abstraction layer.

How does ERPClaw stay portable between SQLite and PostgreSQL?

PyPika constructs queries in a dialect-aware way, and the connection layer hands the application a connection object that honors the same interface regardless of backend. Action handlers do not know which engine is underneath. Schema DDL is the only place where the two backends diverge meaningfully, and that divergence is mechanical translation, not redesign.

What happens when I outgrow SQLite?

You switch the backend to PostgreSQL. The query layer is dialect-aware, the schema migrates with mechanical translation, and the application code is unchanged. The trigger condition is sustained write contention or a hard requirement that SQLite cannot satisfy (replication, multi-writer, regulatory tooling), not raw size. SQLite databases of tens of gigabytes are common in production.

How do you back up a SQLite database that is being written to?

Use the SQLite backup API, or rely on WAL mode and copy the main file plus the `-wal` and `-shm` sidecars. ERPClaw’s recommended backup script does both: a hot snapshot on a schedule, plus an off-host copy via `litestream` for disaster recovery. For PostgreSQL deployments, the standard `pg_basebackup` plus WAL archiving stack applies.

Why not DuckDB?

DuckDB is excellent for analytical workloads, and ERPClaw will likely use it for reporting against historical data. For OLTP (the transactional submit path that posts to the GL), SQLite is the default and PostgreSQL is the alternative. DuckDB is complementary to both, not a replacement.

Does SQLite support JSON columns?

Yes, since 3.38 it has the JSONB-equivalent functions, and from 3.45 it has true binary JSON storage. ERPClaw uses JSON columns for flexible audit metadata; the rest of the schema is normalized. PostgreSQL’s JSONB is more capable, and ERPClaw uses the richer feature set when the backend is PostgreSQL.

How do you handle schema migrations without `ALTER TABLE` limitations?

SQLite’s `ALTER TABLE` is more limited than PostgreSQL’s, but the workarounds (create new table, copy data, drop old, rename) are well understood and easy to script. ERPClaw’s `init_db.py` per module owns the schema for that module’s tables, and migrations are versioned alongside the module. On the PostgreSQL backend, migrations use direct `ALTER TABLE` where possible.

## Closing

Picking SQLite as the default is not a contrarian flex, and it is not an exclusion of PostgreSQL. It is what happens when you let the workload pick the database instead of the other way around, and then build the abstraction so the workload can change its mind. For a self-hosted, AI-native ERP that targets the 40 to 500 employee mid-market, SQLite is the right default for the same reason a single binary is the right deployment target: simplicity compounds when the operator is not a specialist. For the deployments that need PostgreSQL, the backend is there, fully supported, and a config change away.

If you want to see the architecture in practice, the ERPClaw codebase is open source and the entire database layer (PyPika abstraction, pragmas, transaction boundaries, GL validation pipeline) is in the open. The pricing page is short for the self-hosted edition, because the answer is zero. If you have a workload you think breaks the SQLite default, run the PostgreSQL backend; that is what it is for.

---

## The State of AI in Accounting: A 2026 Outlook
URL: https://www.erpclaw.ai/blog/state-of-ai-accounting-2026/

# The State of AI in Accounting: A 2026 Outlook
Where AI actually shipped value in accounting by 2026, what underperformed the hype, the automation tooling stack that earns its place, and what comes next.

The 2018 prediction was that AI would transform accounting by 2025. The reality in 2026 is more granular: AI shipped real value in three specific workflows, made noise in three more without much value, and the cloud-accounting category is starting to bifurcate into AI-native and AI-decorated camps that have different addressable markets.

This post is a state-of-the-category outlook. Where AI actually moved the work, where it didn’t, the automation tooling stack that earns its place, and where the next 12-24 months are heading.

## Where AI actually shipped value

The pattern is consistent: AI works where the task is structured, repeated, and tolerant of imperfect output that a human can correct.

Transaction categorization at scale. This was the obvious first beachhead and the one with the most measurable impact. Rules-based engines plateau: vendor names drift, memo text varies, new vendors keep arriving. Modern AI-driven systems learn from the business’s own correction history, so accuracy compounds month over month instead of plateauing. For a business with thousands of monthly transactions, that compounding is the difference between fixing a large share of categorizations by hand and fixing a small one.

Anomaly detection. A model trained on the business’s normal transaction patterns flags structural weirdness for human review. The flags are noisy at first and tune over six months. The value is finding errors and fraud earlier than the manual review cycle would, sometimes weeks earlier, occasionally months. Most of the larger AP fraud cases prevented in the last three years involved AI flagging the first transactions in a pattern before the pattern grew.

Reconciliation drafting. Bank reconciliation, intercompany reconciliation, payout reconciliation against payment processors. AI proposes matches, drafts adjusting entries for the mismatches, surfaces the items that need human judgment. The hard work (deciding which side is right when both sources disagree) stays with the human; the rest gets done.

All three are at-scale problems where AI replaces high-volume judgment work. The tool earned its place.

## Where AI made noise but not value

Three areas that vendors marketed heavily and customers did not adopt.

AI-generated financial commentary. Most accounting platforms added a “explain this report” feature. The output reads like financial commentary; it does not contain the context that makes financial commentary useful (knowledge of the business’s operational decisions in the period, of macro context, of forward plans). Controllers tried it, found the output generic, and went back to writing the commentary themselves. Some platforms have iterated and the output is now usable as a first-draft starter, but the willingness-to-pay has not materialized.

Conversational accounting. “Ask your books a question” chat interfaces. Customers tried them; the answers were correct on simple queries and wrong or vague on complex ones. The trust threshold for accounting data is high; one wrong answer in five queries kills usage. Most adoption is for very narrow lookups (“what was my Q1 revenue”) rather than the analytical questions vendors marketed.

AI tax preparation. The pitch was that AI would file simple business tax returns end-to-end. The reality is that simple business tax returns are not as simple as the pitch assumed, and the regulatory liability of the wrong-thing-signed is high enough that no vendor has actually shipped end-to-end AI filing for any meaningful customer segment. The work that did ship is AI-assisted preparation with human sign-off, which is helpful but is the same shape as everything else above.

## The automation tooling stack of 2026

The stack that actually shows up in production finance ops has four layers.

### Layer 1: Connectivity

Bank feeds, payment processor feeds, e-commerce platform feeds, ERP/CRM/operational tool integrations. Plaid, Yodlee, Codat, Rutter, and direct vendor connections form the data layer. This is unglamorous infrastructure; it is also where most automation projects fail when the integration breaks at month-end.

Mature 2026 stacks have monitoring on the connectivity layer: alerts when a feed stops syncing, automatic retry on failures, dashboards on data freshness. The stacks that skip this discover the gap in production.

### Layer 2: Workflow orchestration

Zapier, Make, n8n, and increasingly the orchestration layer built into modern AI-native ERPs (ERPClaw included). This layer handles the cross-tool workflows: when an invoice gets approved in tool A, post the journal entry in tool B, notify the customer in tool C, update the project record in tool D.

The 2018-era version of this was RPA: brittle scripts that simulated user clicks. The 2026 version is API-driven workflows with conditional logic and human-in-the-loop gates where appropriate. RPA still exists for tools that have not modernized their APIs but the share is declining.

### Layer 3: AI services

Three categories. Category-detection and anomaly-detection models trained on each business’s history (the workhorses). LLM-based services for draft generation, summary writing, and customer correspondence (useful but supervised). Vertical AI services for specific tasks (expense receipt OCR + categorization, contract review for billing terms, sales tax determination).

Most of the AI services are best-of-breed point solutions that integrate with the accounting platform rather than being native to it. Vendors who tried to build everything in-house have generally fallen behind specialists.

### Layer 4: Approval and audit

The least-discussed but most-important layer for production finance. Human-in-the-loop approval queues, role-based routing, audit-trail logging that captures the AI’s recommendation alongside the human override. This layer is what makes the rest defensible during audits and during incident reviews.

The stacks that skip this layer ship faster but cannot defend their controls to auditors. Public companies cannot use them; private companies that intend to ever raise institutional capital cannot use them; even closely-held businesses regret the gap the first time they have to reconstruct a decision trail.

## What’s next: from co-pilot to operator

The 2022-2025 era was the co-pilot era: AI assists a human accountant. The next era, already visible in 2026, is the operator era: AI handles the end-to-end workflow with humans on exceptions.

This is happening fastest in transactional accounting (AP, AR, expense management, basic categorization) where the decisions are bounded and the audit trail is clear. It is happening slower in judgment-heavy accounting (revenue recognition, complex tax, valuation) where regulatory accountability still rests on the licensed human.

For AI-native ERPs (the 5 that earn the label), the operator pattern is built in: the AI agent is the primary user; the accountant supervises and signs. For AI-decorated systems, the operator pattern would require an architectural rebuild that the vendors have not committed to.

The market implication: the gap between AI-native and AI-decorated widens through 2027-2028. Customers who are happy with current Xero/QuickBooks functionality stay; customers who want the operator pattern migrate. The size of each group is hard to estimate but the migration cohort is real and growing.

## What this means for SMB owners

Three practical implications.

If your current accounting tool is working, stay. AI in accounting is real but the value lives in specific workflows. If you do not have those workflows in volume, the value is limited.

If you are evaluating new tools, evaluate against the four-layer stack. Vendor demos focus on Layer 3 (the AI services) because they are the most impressive. The other three layers matter at least as much. Ask about connectivity reliability, workflow orchestration patterns, and audit trail capture.

If you are at the operator-pattern transition point (you have bounded, repeated, high-volume accounting work where you have been doing manual review for AI suggestions and the suggestions are mostly right), evaluate the AI-native ERPs. The migration is not trivial but the operator pattern compounds; the businesses that get there first capture the productivity gains earlier.

## Closing

AI in accounting did what AI usually does in mature industries: shipped value in specific workflows, underperformed in the ones that were always going to be harder, and is now bifurcating the category between products that committed to the new pattern and products that decorated the old one.

The 2018 prediction was right about the direction and wrong about the timeline. The 2026 reality is partial, specific, and uneven across the stack. For SMB owners, the practical move is to evaluate against the four-layer stack, prioritize the workflows where AI actually delivers value, and treat the rest as marketing for now.

The next 12-24 months will be the operator-pattern era getting traction in the segments that can absorb it. Whether that includes your business is a question of workflow shape, not of how excited the vendors sound.

Tagsai-accountingtrendsautomationvision
Share:XLinkedInHacker News

---

## Stripe ASC 606 Revenue Recognition: A Founder's Guide
URL: https://www.erpclaw.ai/blog/stripe-asc-606-revenue-recognition-guide/

# Stripe ASC 606 Revenue Recognition: A Founder's Guide
ASC 606 for SaaS founders on Stripe: the five-step model, journal entries, mid-cycle changes, refunds, trials, and an audit-prep checklist.

ASC 606 came into force for private companies in 2018. For about three
years after, most SaaS founders cheerfully ignored it. They booked
Stripe deposits as revenue the month the cash hit the bank and went
back to building features.

Then somebody asks for audited financials. A potential acquirer. A
Series A lead. A new bank line. And the same conversation happens:

“We need to restate three years of revenue.”

Three weeks of panic, a five-figure invoice from a Big Four firm, and
a brand new column in the cap table called “audit findings.” It
happens because nobody told the founder that “Stripe deposited $14,200
in March” is not the same sentence as “we earned $14,200 of revenue in
March.”

This post is the practical guide I wish someone had handed me when I
was wiring up my first SaaS company’s books. I spent several years as
an SAP and ASC 606 implementation lead at Accenture before founding
ERPClaw, so I have watched a lot of these conversations from both
sides of the table. We will walk through the five-step ASC 606
framework using a concrete Stripe subscription, write the actual
journal entries, work through the awkward edge cases, and finish with
a checklist for your first audit.

This post mixes founders and CPAs. If you already know what a
performance obligation is, skip the definitions in brackets. If
you already know what MRR means, the same applies in the other
direction.

## What ASC 606 actually says

ASC 606 is the revenue recognition standard issued by FASB
and IASB jointly. “ASC 606” is the FASB codification; “IFRS 15” is the
international equivalent. Same standard, different name.

The core idea is one sentence: revenue should be recognized when (or
as) the seller transfers control of the promised good or service to
the customer. Cash receipt is irrelevant. Invoice date is
irrelevant. Contract signature date is irrelevant. What matters is
when the customer actually gets what they paid for.

For a SaaS subscription, “what they paid for” is access to your
software for some period. If they paid for twelve months in January,
you have not earned that money in January; you earn it day-by-day
across the year. The cash sits on your balance sheet as a liability
called deferred revenue (sometimes “unearned revenue”: money the
customer paid you that you have not yet earned because you still owe
service). It stays there until you deliver.

That is the entire intuition. Everything else is mechanics.

## The five-step framework, applied to a Stripe subscription

The standard prescribes a five-step model for any contract. For a
straightforward SaaS subscription on Stripe, the steps map cleanly.
Let us use a running example: Acme Inc subscribes to your $1,200/year
“Pro” plan on January 1, 2026, paid up front via Stripe.

### Step 1: Identify the contract

A contract creates enforceable rights and obligations. For SaaS, it is
usually three artifacts: your Terms of Service, your pricing page, and
the Stripe subscription record itself.

Practically: a Stripe subscription is the contract. When Acme
clicks Subscribe and Stripe creates a `sub_xxx` record, you have a
contract. The subscription ID is your identifier;
`current_period_start` and `current_period_end` are the boundaries;
`plan_amount` and `plan_interval` define the consideration.

If you sell enterprise contracts with a separately-signed MSA, the MSA
is the contract and Stripe is just the billing mechanism. Same
accounting; the artifact you point an auditor at is different.

### Step 2: Identify performance obligations

A performance obligation is a distinct promise to transfer a good
or service. (For founders: the “thing you are on the hook to deliver”
line item.) For a typical single-product SaaS sub, there is one:
deliver software access for the subscription period.

Trivial for most plans. It gets interesting when your plan bundles
things:

- SaaS access plus implementation services = two obligations

- SaaS access plus training = potentially two obligations (depends on
whether training is “distinct”)

- SaaS access plus a hardware kit = two obligations

- A “platform fee” plus a “per-transaction fee” = usually one combined
obligation, recognized as transactions occur

The test for “distinct” is whether the customer could benefit from
each on its own. A training course you only sell to subscribers is
probably not distinct. A training course you also sell standalone is.

For the rest of this post assume one obligation. If you have
multiple, repeat steps 3 through 5 per obligation, which is where this
stops being a spreadsheet exercise and starts being a software one.

### Step 3: Determine the transaction price

The amount you expect to be entitled to in exchange for delivering.
For Acme: $1,200, paid up front.

Adjustments that affect transaction price:

- Discounts and coupons: reduce the price. If Acme used 20% off
and paid $960, transaction price is $960.

- Expected refunds: estimated returns reduce price. For SaaS,
small but not zero (think 7-day money-back).

- Variable consideration: usage pricing, performance bonuses,
success fees. Estimate the expected amount, constrained so you do
not recognize revenue you might have to reverse.

- Significant financing component: prepayments over a year should
technically split out interest. Most SaaS companies treat this as
immaterial for twelve months or less.

Transaction price for our example: $1,200.

### Step 4: Allocate the transaction price

Allocation only matters when you have multiple performance obligations.
For Acme’s single-obligation subscription, the entire $1,200 is
allocated to the SaaS access obligation.

If you bundled a $200 implementation fee into the $1,200 price, you
would allocate based on standalone selling prices (SSP: what each
component would sell for if you sold it alone). Say SaaS access alone
is $1,200/year and implementation alone is $300:

- Total SSP = $1,500

- Allocate $1,200 × ($1,200 / $1,500) = $960 to SaaS access

- Allocate $1,200 × ($300 / $1,500) = $240 to implementation

This is “relative SSP allocation” and it is where most DIY ASC 606
spreadsheets quietly start lying.

### Step 5: Recognize revenue when (or as) the obligation is satisfied

For Acme: satisfied over time, ratably across twelve months. So
you recognize $1,200 / 12 = $100/month.

For one-time obligations (hardware delivery, implementation milestone),
recognize at the point in time the obligation is satisfied. For
variable obligations (transaction-based pricing), recognize as the
customer consumes.

That is the whole framework. Now the journal entries.

## The journal entries you would actually post

Every example below uses debits and credits from your perspective
(the SaaS company). Stripe takes 2.9% plus $0.30 per card transaction;
we will use a flat $35 fee on the $1,200 charge for arithmetic
clarity (real number is $34.80; close enough).

### Entry 1: Acme pays $1,200 on January 1

When Stripe confirms the charge succeeds:

| Account
| Debit
| Credit

| Stripe Clearing (asset)
| $1,165
|

| Stripe Fees (expense)
| $35
|

| Deferred Revenue (liability)
|
| $1,200

A few things worth noting:

- Stripe Clearing is an asset account that holds funds Stripe owes
you but has not paid out yet. It typically takes 2 to 7 business
days for Stripe to deposit into your bank.

- Stripe Fees are recognized immediately as an expense. They are
not deferred over the subscription period (this is one of the rare
cases where cash-basis intuition is correct).

- Deferred Revenue is a liability. You owe Acme twelve months of
software access; until you deliver, that obligation lives on the
balance sheet.

- Subscription Revenue is zero at this point. You have not
earned anything yet.

### Entry 2: Stripe pays you out (a few days later)

| Account
| Debit
| Credit

| Bank (asset)
| $1,165
|

| Stripe Clearing (asset)
|
| $1,165

Stripe Clearing nets to zero for this transaction. This is the
“three-layer reconciliation” pattern: every payout from Stripe must
match the net of the underlying transactions, or your books are wrong.

### Entry 3: End of January, recognize 1/12

| Account
| Debit
| Credit

| Deferred Revenue (liability)
| $100
|

| Subscription Revenue (income)
|
| $100

You have now delivered one month of access. Deferred Revenue drops by
$100; Subscription Revenue picks it up.

Repeat this entry every month. By December 31, Deferred Revenue for
Acme’s contract is $0 and you have recognized $1,200 of revenue
across the year.

### Entry 4: Acme cancels at end of April with a pro-rata refund

Acme used four months ($400 earned) and wants a refund for the unused
eight months ($800).

The refund settles in two parts. The cash refund:

| Account
| Debit
| Credit

| Deferred Revenue (liability)
| $800
|

| Stripe Clearing (asset)
|
| $800

You no longer owe Acme the eight months of access (because you are
giving the money back), so Deferred Revenue is cleared. You also no
longer have $800 sitting in Stripe Clearing waiting to be paid out;
Stripe pulls it back to refund Acme.

Note: Stripe does not refund the original processing fee. The $35
you paid in January stays gone. Some Stripe regions have changed this
policy over the years; check your account, but assume the worst.

### Entry 5: Acme cancels at end of April but elects “no refund, end-of-period grace”

Same scenario, but Acme just stops using the product on April 30 and
your terms say “no refund.” The eight months of unrecognized revenue
are still your money, and you have just been told you no longer need
to deliver anything.

| Account
| Debit
| Credit

| Deferred Revenue (liability)
| $800
|

| Subscription Revenue (income)
|
| $800

The full $800 hits revenue at the moment of cancellation, because the
performance obligation is discharged. Whether by delivery or by
extinguishment, it is gone, and the deferred liability has to clear.

This is a place where DIY spreadsheet bookkeeping silently goes wrong:
founders forget that “customer cancelled, kept the money” is a revenue
recognition event, not a non-event. The auditor will catch it.

## The awkward stuff: mid-cycle changes

Real SaaS contracts do not sit politely from January to December. They
get upgraded, downgraded, paused, and modified. Stripe handles the
billing math correctly with prorations; ASC 606 has its own opinion on
how those prorations should hit your books.

### Mid-cycle upgrade

Acme upgrades from $100/month to $200/month on day 15 of a 30-day
cycle. Stripe creates a proration invoice for the additional $50 (half
a month at the $100 incremental price), charged immediately.

The recognition schedule splits:

- Days 1 to 15: recognized at $100/month rate, so $50 earned. No
change.

- Days 16 to 30: $200/month rate, so $100 earned across the remainder.

- The $50 proration invoice goes to Deferred Revenue first, then
recognizes ratably across days 16 to 30.

End-of-month total recognized: $50 plus $100 = $150. Cash collected:
$100 plus $50 = $150. Books balance.

If you do this in a spreadsheet, you will mess it up the third time it
happens. Trust me.

### Mid-cycle downgrade

Acme downgrades from $200/month to $100/month on day 15. Stripe issues
a credit note for the unused half-month at the higher tier ($50
credit). The credit applies to the next invoice rather than refunding
cash.

Journal entries:

- Reverse the $50 of unearned high-tier revenue:
`DR Deferred Revenue $50, CR Stripe Clearing $50` (credit memo).

- The credit sits in Stripe as a customer balance, applied at next
invoice.

The recognition schedule for days 16 to 30 reverts to the $100/month
rate.

### Annual subscription with a discount

Acme subscribes to the $1,200 annual plan with a 20% off coupon and
pays $960 up front.

The transaction price (Step 3) is $960, not $1,200. The discount
is not a separate accounting line item; it is absorbed into the price
and disappears.

| Account
| Debit
| Credit

| Stripe Clearing (asset)
| $933 (after $27 fee)
|

| Stripe Fees (expense)
| $27
|

| Deferred Revenue (liability)
|
| $960

Monthly recognition: $960 / 12 = $80/month, not $100.

The fact that the “list price” was $1,200 is irrelevant for revenue
purposes. It might matter for sales reporting or pricing analytics,
but not for GAAP.

### Trial-to-paid transition

Acme signs up for a 14-day free trial on January 1. Stripe creates the
subscription in `trialing` status; no charge, no payout, no journal
entry. Deferred Revenue is zero, because nothing has been
collected and nothing has been promised in exchange for cash.

On January 15, the trial ends. Stripe attempts the first charge, it
succeeds, and the regular flow kicks in: $1,200 to Stripe Clearing,
$1,200 to Deferred Revenue.

If Acme cancels during trial, no journal entries are ever posted. The
fact that the trial existed has no GL consequences. (For metrics, your
MRR/ARR system might track trial conversion rates, but that is
analytics, not accounting.)

### Failed payment then recovery

Stripe attempts to charge Acme on day 1 of a renewal. The card is
declined. Stripe retries on days 3, 5, and 7, and the charge finally
succeeds on day 7.

You post one journal entry, dated day 7, for the successful
charge. The four failed attempts have no GL impact. Log them for audit
(a proper integration syncs them to a `stripe_charge` table with
`status='failed'`), but they do not touch the GL.

If the customer is in `past_due` status with a recognition schedule
that should have started on day 1, decide policy: backdate recognition
to day 1 (when access was nominally granted) or day 7 (when payment
arrived)? Most SaaS companies pick day 7. No payment, no obligation,
no recognition.

## When you can DIY and when you cannot

- Monthly-only, no annuals, under $1M ARR: you probably do not
need ASC 606 in practice. Recognize cash as revenue, document the
policy, move on.

- Annual contracts, pre-audit: DIY in a spreadsheet is feasible.
The math is grade-school arithmetic. The risk is consistency, you
forgetting to update the schedule when contracts change.

- Annual contracts plus a Series A or acquisition on the horizon:
get this right now. Retroactive rebuilds of three years of
schedules cost 10x to 100x what real-time would have.

- Multi-element bundles, Connect platform fees, variable pricing:
get a CPA who knows ASC 606. Steps 4 and 5 involve real judgment.

- Public-company-bound: not a question. Software, controller, Big
Four sign-off.

## Audit-prep checklist

When the auditor shows up they will ask for evidence. Minimum viable
file:

- Revenue recognition policy memo, one page in plain English:
subscriptions recognized ratably from access-grant date,
implementation at milestone, Stripe fees expensed as incurred,
refunds reverse unrecognized portion (or hit revenue if already
recognized).

- Per-contract recognition schedule: for every active sub, a row
with contract value, start, end, monthly recognition, and
recognized-to-date.

- Period-end recognition journal entries: one voucher per
month-end cycle, with per-contract detail attached.

- Deferred Revenue rollforward: opening + new + recognized -
refunds = closing. Closing must reconcile to both the GL liability
balance and the sum of unrecognized contract values.

- Stripe payout reconciliation: every payout matched to bank
deposits, with underlying charges, fees, and refunds tied out.
Stripe Clearing must net to zero each period, or the discrepancy
must be explained.

- Modification log: every upgrade, downgrade, mid-cycle change,
and the journal entry that recorded it.

- Refund register: every refund, with original charge reference
and treatment (reversed deferred vs. booked against revenue).

If you can produce these seven artifacts on demand, you will pass an
ASC 606 audit. If you cannot, you will spend a lot of money having
your auditor produce them for you.

## Tools that can help

Ordered roughly by founder budget:

- Google Sheets. Free. Works for 50 or fewer active subs if you
are disciplined. Breaks the moment you forget to update for a
contract change. The number-one cause of audit findings I have seen
is “we maintained recognition in a spreadsheet that diverged from
the GL.” That said, if you are a five-customer company with three
annual contracts, a sheet is the right answer for now.

- Stripe’s Revenue Recognition product.
$0.04 per processed transaction. Computes schedules from Stripe data
and gives you journal entries to post into your accounting system.
Good if Stripe is your sole source of truth and you are fine
hand-posting entries into QuickBooks or Xero each month.

- Synder, Bookkeep, A2X: connector tools, $20 to $100/month,
mostly ecommerce-focused. ASC 606 specifically is light, but they
sync the underlying charges and fees cleanly into your GL.

- Rillet. SaaS-focused finance platform with strong ASC 606
support. Enterprise pricing, typically $1K to $3K/month. See our
Rillet comparison.

- Sage Intacct. Legacy mid-market, $400 to $1,500/month per user,
full ASC 606 module. Overkill for most early-stage SaaS but the
default landing spot once a controller arrives.

- NetSuite. If you are asking whether you need NetSuite, you do
not.

- ERPClaw. Free, open-source, AI-native, self-hosted on your own infrastructure.
Implements the five-step model for Stripe
subscriptions and posts journal entries with full audit trail. The
other tools above are all AI-decorated; ERPClaw is the only AI-native
option in the category, which is part of why the price can be zero.
See the Stripe integration page. We built it
because we wanted what Rillet does at zero monthly cost.

Pick the cheapest one that produces the seven audit artifacts above
without you having to think about it. For most pre-Series-A SaaS
companies, Stripe’s own Revenue Recognition product plus a tidy
QuickBooks file is enough.

## Frequently asked questions

### Do I need ASC 606 if I am cash-basis for tax purposes?

Tax basis and book basis are separate. You can file taxes on a cash
basis and still need ASC 606 for your audited financials, board
reporting, or due diligence. Most VC-backed SaaS companies maintain
two sets of books for exactly this reason.

### What if I just recognize annual prepayments evenly across 12 months, is that good enough?

For a single-product, single-period subscription with no mid-cycle
changes and no refunds, yes. Straight-line monthly recognition is
correct for SaaS access. The complications come from refunds,
upgrades, downgrades, trials, and bundles, exactly the things that
happen in a real customer base.

### My subscriptions are monthly, not annual. Do I still need to defer?

If a customer pays on January 1 for January access, recognize the full
amount in January. In practice you can treat monthly subscriptions as
recognized in the period charged, provided your billing dates align
with your accounting periods.

If a customer pays on January 20 for service through February 19, you
should split: roughly 36% in January, roughly 64% in February. Most
companies ignore this because the error washes out across periods, but
it is technically incorrect.

### What about usage-based or metered billing?

Recognize as the customer consumes. If Acme uses 10,000 API calls in
March and you bill $0.001 per call, you recognize $10 of revenue in
March. Stripe’s metered billing handles the invoicing; you just need
to make sure your accounting picks up the right period.

### How do I handle Stripe processing fees?

Expense them in the period incurred. They are not part of revenue
recognition and they are not refundable when you refund a customer.
Treat them like any other vendor fee.

### What about the discount on an annual plan, do I recognize the “list price” or the discounted price?

Discounted price. The transaction price under ASC 606 is what the
customer is contractually obligated to pay, net of discounts. The
“list price” is a marketing concept with no GAAP relevance.

### How does ASC 606 treat Stripe Connect application fees?

Application fees you collect from connected accounts are your revenue,
recognized when the underlying transaction completes (usually
point-in-time). The gross transaction is not your revenue if you are
not the principal. Principal-vs-agent is the biggest judgment call in
Connect accounting and worth a CPA conversation if material.

### Will my auditor accept journal entries posted automatically by software?

Yes, provided the software has an audit trail (who, what, when, why)
and the recognition logic is documented. Any reasonable accounting
system, including ERPClaw, provides this.
Auditors care that the logic is documented and reproducible, not who
typed the entry.

## Closing

ASC 606 is not hard. It is fiddly. The five-step framework is
straightforward, the journal entries are standard, and the principles
are intuitive once you internalize that “cash received” and “revenue
earned” are different events.

What makes it hard in practice is consistency: applying the same logic
to every contract, modification, refund, and period, forever. Humans
are bad at this. Spreadsheets are okay until they are not.

If you take one thing from this post: stop booking Stripe deposits as
monthly revenue if any of your subs are annual. Set up a deferral
schedule and recognize ratably. Your future audited self will thank
you.

If you are comparing tools, the
ERPClaw plus Stripe launch post,
the canonical ASC 606 reference, and the
Rillet comparison cover the rest. The
SaaS landing page and pricing round it out.

Questions or edge cases? Email `support@avansaber.com`.

---

## Stripe Connect Application Fees: Accounting and Journal Entries
URL: https://www.erpclaw.ai/blog/stripe-connect-application-fees-accounting/

# Stripe Connect Application Fees: Accounting and Journal Entries
Accounting for Stripe Connect application fees in direct charges, destination charges, and separate charges and transfers, with worked journal entries and refund entries.

It is 11pm and you are six tabs deep into the
Stripe Connect documentation. You
have a marketplace launching in three weeks. Every page seems to
introduce a new concept: `application_fee_amount`,
`transfer_data.destination`, `on_behalf_of`, separate charges and
transfers, reverse transfers.

Underneath the API choice is an accounting choice. Each Connect
pattern produces a different set of journal entries, splits revenue
between your platform books and the connected-account books in
different ways, and treats refunds differently when something goes
wrong. The docs cover it, but in the order an integration engineer
would want it, not in the order a founder designing the business
needs it.

I have implemented all three Connect patterns in production. I spent
several years as an SAP and ASC 606 implementation lead at Accenture
before founding ERPClaw, so I have seen what happens when the
accounting decisions get made by accident rather than on purpose.
This is the explainer I wish someone had handed me the first time.

This guide owns the accounting analysis and worked journal entries. For API fields, ERPClaw configuration, events, and operational flow, use the Stripe Connect fields and flows reference.

We will walk the same $100 customer charge through each pattern, write
the journal entries on both sets of books, work through refunds (the
part that gets people in trouble), and finish with a decision tree.

## What Stripe Connect actually does

Stripe Connect lets your platform process payments and route money to
other Stripe accounts you have onboarded. Those accounts are called
connected accounts: sellers on your marketplace, vendors on your
B2B platform, drivers on your delivery app.

The plumbing question is: when a customer pays $100 on your platform
and $90 is supposed to go to a connected account, whose Stripe
balance does the charge land in first, and how does the connected
account get its share?

Stripe gives you three answers, and they are not interchangeable. The
shape of the money flow drives who is the merchant of record, whose
tax ID is on the receipt, who absorbs chargebacks, and how your
ledger has to look.

The three patterns:

- Direct charges: charge happens on the connected account, your
platform takes an `application_fee_amount` cut.

- Destination charges: charge happens on your platform, you set
`transfer_data.destination` and Stripe immediately moves the
connected account’s share.

- Separate charges and transfers: charge happens on your
platform, then you fire `transfers.create` separately, whenever
you want.

Let us put $100 through each.

## Pattern 1: Direct charges with application_fee_amount

The customer pays the connected account. Your platform is invisible
to the receipt. In this worked example, the connected account is
configured to pay Stripe fees. Stripe debits that account’s balance
for the processing fee, then deducts your `application_fee_amount`
and routes it to your platform balance.

A $100 charge with a $10 application fee and Stripe’s standard 2.9%
plus $0.30 domestic card fee (call it $3.20):

- Customer is charged: $100

- Stripe processing fee debited from connected account: $3.20

- Application fee transferred to platform: $10

- Net to connected account balance: $86.80

### Journal entries on the connected account’s books

| Account
| Debit
| Credit

| Stripe Clearing
| $86.80
|

| Stripe Fees
| $3.20
|

| Application Fee Expense
| $10.00
|

| Revenue
|
| $100.00

The connected account is the principal. They booked $100 of revenue.
They paid Stripe $3.20 and they paid your platform $10 as a
marketplace fee, which is an operating expense on their books.

### Journal entries on your platform’s books

| Account
| Debit
| Credit

| Stripe Clearing
| $10.00
|

| Marketplace Fee Revenue
|
| $10.00

Your platform recognizes $10 of marketplace fee revenue. You never
touched the gross $100. From an
ASC 606 principal-versus-agent
perspective you are clearly the agent here, and your revenue is the
fee you charged for matching the buyer and seller.

This is the Etsy pattern. Each seller is the merchant of record,
each seller has their own Stripe account, the buyer’s receipt shows
the seller’s name. Your platform is a fee-collecting matchmaker.

## Pattern 2: Destination charges with transfer_data

The customer pays your platform. The receipt shows your name. Stripe
charges the customer, then immediately transfers the connected
account’s share based on `transfer_data.destination` and
`transfer_data.amount`.

Same $100 charge, $90 going to the connected account, $10 you keep.
Stripe processing fee comes off your platform side:

- Customer is charged: $100

- Stripe processing fee debited from platform: $3.20

- Transfer to connected account: $90

- Net to platform balance: $6.80

### Journal entries on your platform’s books

The charge:

| Account
| Debit
| Credit

| Stripe Clearing
| $96.80
|

| Stripe Fees
| $3.20
|

| Revenue
|
| $100.00

Then the immediate transfer:

| Account
| Debit
| Credit

| Connect Transfer Expense
| $90.00
|

| Stripe Clearing
|
| $90.00

Net effect: Stripe Clearing rises by $6.80, you booked $100 of
revenue, you booked $90 of cost. Gross margin on the transaction is
$6.80 minus any other COGS.

Whether “Connect Transfer Expense” is a true expense or a contra-revenue
line is the principal-versus-agent question. If the connected account
is delivering the service and you are facilitating, you are the agent
and the $90 should reduce revenue, not appear as an expense. If you
are reselling and the connected account is your supplier, you are
the principal and $90 is COGS.

### Journal entries on the connected account’s books

| Account
| Debit
| Credit

| Stripe Clearing
| $90.00
|

| Revenue
|
| $90.00

The connected account books $90 of revenue, sees $90 hit their Stripe
balance, never sees the gross $100, never sees the processing fee.
They get a clean payable from your platform.

This is the SaaS marketplace pattern. Your platform is the
merchant of record, your tax ID is on the receipt, you handle
customer service. The connected account is essentially a payee that
delivers part of the value. Substack, Patreon, and most “platform
takes a cut” SaaS marketplaces use this.

## Pattern 3: Separate charges and transfers

You charge the customer on your platform with no `transfer_data`.
The full $96.80 (after Stripe’s fee) lands in your platform balance.
Then, whenever you want, you call `transfers.create` to move the
connected account’s share.

This is the escrow pattern. You can hold the money for hours,
days, or weeks. You can release it on a delivery confirmation, a
shipment scan, the end of a return window, or a manual approval. You
can also split the same charge across multiple connected accounts
later, which neither of the other patterns supports cleanly.

### Journal entries on your platform’s books

The charge:

| Account
| Debit
| Credit

| Stripe Clearing
| $96.80
|

| Stripe Fees
| $3.20
|

| Revenue
|
| $100.00

When you decide to release, post a transfer-pending entry:

| Account
| Debit
| Credit

| Connect Transfer Pending
| $90.00
|

| Stripe Clearing
|
| $90.00

When the transfer settles and the connected account confirms receipt:

| Account
| Debit
| Credit

| Connect Transfer Expense
| $90.00
|

| Connect Transfer Pending
|
| $90.00

### Journal entries on the connected account’s books

| Account
| Debit
| Credit

| Stripe Clearing
| $90.00
|

| Revenue
|
| $90.00

Same end state as the destination pattern, but with a holding period
on the platform side. The “Connect Transfer Pending” account is what
your auditor will ask about: it is essentially funds you are holding
on behalf of someone else, which has implications for licensing in
some jurisdictions (look up “money transmitter” before you build a
30-day escrow on a US platform).

This is the pattern Uber-style platforms, real-estate booking sites
with cancellation windows, and B2B marketplaces with delivery
verification all reach for. It is the most powerful and the most
complicated.

## Refunds, the part everyone gets wrong

Refund handling is where Connect accounting goes sideways. The
question is always the same: when the customer gets their money back,
who eats it? Stripe gives you flags to control this, but the flags
only matter if your books mirror what Stripe actually did.

### Direct charge refund

The refund happens on the connected account, because the original
charge happened on the connected account. The flag that matters is
`refund_application_fee=true|false`.

If `refund_application_fee=true`, Stripe pulls the $10 application
fee back from your platform balance and returns it to the connected
account. Stripe debits the $100 customer refund from the connected
account balance, and your platform balance drops by $10.

Connected account’s books:

| Account
| Debit
| Credit

| Stripe Clearing
| $10.00
|

| Application Fee Expense
|
| $10.00

The customer refund is a separate event:

| Account
| Debit
| Credit

| Revenue
| $100.00
|

| Stripe Clearing
|
| $100.00

Platform’s books:

| Account
| Debit
| Credit

| Marketplace Fee Revenue
| $10.00
|

| Stripe Clearing
|
| $10.00

If `refund_application_fee=false`, your platform keeps the $10. The
connected account eats the refund alone. No journal entry on your
platform side. This is the contractually-aggressive option and you
should probably tell sellers up front if you are going to do it.

Note: Stripe does not refund the original processing fee in
either case under standard pricing. That $3.20 stays gone. Connected
account absorbs it. See Stripe’s refund and dispute guidance.

### Destination charge refund

The refund happens on your platform. The flag is
`reverse_transfer=true|false`, which controls whether the $90 you
already pushed to the connected account gets pulled back.

If `reverse_transfer=true`, the platform refunds the customer $100
and Stripe pulls the $90 back from the connected account.

Platform’s books, the refund:

| Account
| Debit
| Credit

| Revenue
| $100.00
|

| Stripe Clearing
|
| $100.00

The transfer reversal:

| Account
| Debit
| Credit

| Stripe Clearing
| $90.00
|

| Connect Transfer Expense
|
| $90.00

Connected account’s books:

| Account
| Debit
| Credit

| Revenue
| $90.00
|

| Stripe Clearing
|
| $90.00

Net effect: everybody is back to zero, except the platform still ate
the $3.20 in Stripe fees.

If `reverse_transfer=false`, the platform refunds the customer the
full $100 but the connected account keeps the $90. Your platform is
out $93.20 ($100 refund plus $3.20 fee, minus the $10 you would have
kept). This is the right choice when the connected account already
shipped a physical good and you do not want to chase them for the
money. It is the wrong choice when the customer never received
anything; that money is just gone.

### Separate transfer refund

You reverse the charge first, then optionally call
`transfers.createReversal` to pull the transfer back.

The flexibility is the point. You can refund the customer immediately
and decide later whether to claw back the connected account’s share
based on whatever business rule applies (was the order shipped? did
the seller ship in time? was the dispute the seller’s fault?).

The discipline is also the point. Every refund without a matching
reversal decision creates an accounting question that has to be
answered by hand or by code. Spreadsheet bookkeeping breaks here
within the first month.

### Application-fee-only refund

Stripe also lets you refund only the application fee without
refunding the underlying charge. Useful when your platform overcharged
its marketplace fee or you want to give a connected account a
goodwill credit.

Platform’s books:

| Account
| Debit
| Credit

| Marketplace Fee Revenue
| $5.00
|

| Stripe Clearing
|
| $5.00

The $5 returns to the connected account. The customer is unaffected.

## Two sets of books, one ledger system

Every example above had separate journal entries on the platform’s
books and on the connected account’s books. That is not stylistic.
Each Stripe account is a separate accounting entity. The
connected account has its own legal identity, its own 1099-K, its
own tax filings.

If your platform also keeps the books for your sellers (common for
SaaS marketplaces and embedded-finance plays), you need your
accounting system to keep those ledgers separate. One general ledger
per Stripe account, with its own chart of accounts and its own
period-end close. Consolidated reporting on top, but the source
ledgers do not mix.

Most off-the-shelf SaaS accounting tools assume one entity, which
works fine for the platform alone. The moment you need to also track
connected-account books for reporting or compliance, multi-entity
ledger support is the difference between a tool that handles Connect
properly and one that stuffs everything into a single chart of
accounts and labels rows with a metadata field.

ERPClaw treats every Stripe account as a separate `stripe_account`
row with its own GL account namespace. The
Connect platform fees reference
has the full schema.

## Picking the right pattern

The choice is business model, not technical preference.

Use direct charges when: each connected account is a separate
brand, has its own customer relationship, and wants the receipt to
show their name. Etsy/Airbnb style. Connected accounts file their
own 1099-Ks. Your platform is a fee collector and never sees the
gross transaction on its books.

Use destination charges when: your platform is the merchant of
record, your name is on the receipt, and you handle customer
service, refunds, and chargebacks centrally. SaaS marketplaces where
the platform sells and splits revenue with vendors (Substack,
Patreon, course platforms). Settlement is immediate.

Use separate charges and transfers when: you need to hold money
before releasing it (delivery verification, return windows), split a
single charge across multiple connected accounts, release on a
schedule that is not the charge timestamp, or run an
escrow/hold-and-release model. You take on additional accounting and
possibly licensing complexity.

The trap to avoid: do not mix patterns within the same Connect
platform. Stripe lets you. Your books will hate you for it. Pick
one pattern at design time and stick with it.

## Tools that can automate this

There is a depressing amount of manual work in Connect accounting if
you do not automate it. A few options, ordered by founder budget:

- Stripe’s own reporting.
Free with your account. Connect-aware reports that show
application fees, transfers, reversals, and per-connected-account
balances. The output is CSV/JSON you push into your accounting
system; the math itself is correct.

- Stripe Sigma. SQL on top of your Stripe data, $0.02 per query.
Useful when you have outgrown the canned reports. Still no journal
entries; you build those yourself.

- Synder and similar connectors. Sync individual Connect charges
into QuickBooks or Xero with some Connect awareness. Light
multi-entity support. Good for small platforms on a single
destination-charge pattern.

- ERPClaw. Free, open-source, AI-native, self-hosted on your own infrastructure.
Implements all three Connect patterns, handles per-account ledgers
natively, and posts journal entries with audit trail. See the
Stripe integration page and the
SaaS landing page.

- Sage Intacct and NetSuite. Connect-capable modules through
partners. Real money ($1K to $10K/month) and real implementation
projects. Default landing spot once you have a controller.

Most platforms can get to $5M of GMV on Stripe’s own reporting plus a
careful person and QuickBooks. After that, ledger complexity wins.

## Frequently asked questions

### If I use direct charges, do I have to issue 1099-Ks to my connected accounts?

No. The connected account is the merchant of record on a direct
charge, and Stripe issues the 1099-K to them on their own account.
Your platform issues 1099s only for its own contractor or vendor
payments, not for marketplace gross volume.

### If I use destination charges, who issues the 1099-K?

You do, for the gross transaction, because you are the merchant of
record. You also have the option of issuing 1099-MISC or 1099-NEC to
the connected accounts for the transfers you made to them, depending
on how you classify the relationship. This is one of the conversations
worth having with a CPA before you scale.

### Can I switch patterns later if I picked the wrong one?

Technically yes, business-realistically no. Switching patterns means
new charges flow through different code paths and produce different
journal entries, and your historical reports become apples-and-oranges.
You can do it, but the migration project is real. Pick carefully the
first time.

### What about Stripe Tax?

Different product. Stripe Tax handles US sales tax calculation,
registration tracking, and filing reminders. It does not change the
Connect pattern question. You still have to pick direct, destination,
or separate.

### Do application fees count as my revenue under ASC 606?

Yes, recognized at the point in time the underlying transaction
completes (usually instantly). The principal-versus-agent question
matters for whether you also recognize the gross transaction. Direct
charges = you are clearly the agent, fee is your revenue, gross is
not. Destination charges = depends on facts, see your CPA. The
ASC 606 deep dive
walks through this for SaaS subscriptions; the principle is the same
for Connect fees.

### What happens to application fees during a chargeback?

Same as a refund with `refund_application_fee=true` by default. Stripe
also charges a $15 chargeback fee that comes out of the platform or
the connected account depending on your platform settings. Set the
policy explicitly; the default may not be what you want.

## Closing

Stripe Connect is one of the few API surfaces where the technical
choice and the accounting choice are the same choice. Picking direct
versus destination versus separate is a business question about who
is the merchant of record, who eats the fee, and how you want your
books to look five years from now.

Walk the $100 example through both sets of books for each pattern
before you write any code. The pattern with the cleanest journal
entries is usually the right one for your business.

The Connect platform fees reference,
Stripe integration page, and pricing
cover the rest. The
ASC 606 guide is
the companion read for SaaS-style platforms.

Questions or edge cases? Email `support@avansaber.com`.

---

## Synder Alternative: One Tool Instead of Synder + QBO
URL: https://www.erpclaw.ai/blog/synder-alternative-erpclaw-shopify/

# Synder Alternative: One Tool Instead of Synder + QBO
Synder syncs your Shopify sales into books you still pay for. ERPClaw is the books: free, open source, with the Shopify integration built in.

If you pay for Synder every month, look at the bill underneath it. You are also paying for QuickBooks Online. That is not optional. Synder does not store your books. It reads your Stripe and Shopify activity, formats it, and shoves it into QBO. Without QBO on the other end, Synder has nowhere to put anything.

That is not a bug in Synder. It is the shape of the product. But it is the reason you are writing two checks instead of one. The first check buys the connector. The second check buys the place the connector dumps your data. Add them up and call that number what it really is, which is the cost of an accounting setup, not the cost of a sync tool.

ERPClaw is a different shape. One tool, not two, and it is AI-native, not AI-decorated. The Stripe and Shopify integrations write straight into a real general ledger that lives on your own computer or server. No QBO. No second subscription. No monthly fee at all. open source, source on GitHub, free to install, free to run, free for as long as you want to use it.

Synder is a competent connector (launched in 2017, parent company CloudBusiness founded 2016) with AI features added on top later. ERPClaw is the only AI-native option in this category, which means the assistant is the primary interface and the architecture was designed around AI from the first commit. That is a structural difference no Synder-shaped product can retrofit without throwing the codebase away. The argument is laid out in AI-decorated vs AI-native software if you want the long form.

This is the long version of the Synder vs ERPClaw comparison page, with more reasoning and fewer marketing words. If you are looking at Synder right now and want to know what the alternative actually does, keep reading.

## What Synder genuinely gets right

I want to start here, because Synder is a real product that helps a lot of merchants, and a fair comparison is the only one worth your time.

Synder’s main strength is breadth of channels. They sync Stripe, Shopify, Amazon, eBay, PayPal, Square, Etsy, WooCommerce, and a long list of others. If you sell across five rails and you are tired of CSV exports, Synder pulls all five into one QBO file and the copy-paste stops.

The signup flow is friendly. Connect Stripe, connect Shopify, connect QBO, pick your settings, and you are done in an afternoon. That is rare in accounting software.

They have years of QBO mapping logic baked in. Refunds, processing fees, multi-currency, sales tax. The awkward edge cases that a new tool would get wrong on day one. Synder has seen those cases thousands of times and the mapping is reliable.

So if you sell on five marketplaces and you want everything pushed into QBO, Synder does that job well. They have earned the customers they have.

## Where the Synder bill starts to bite

The trouble with Synder is not the software. It is what the software costs to run, end to end, after you sign up.

### You pay twice, every month, forever

Synder’s plans run from $65 to $275 a month (as of June 2026) depending on tier and transaction volume. Then QuickBooks Online Plus is another $115 a month on top. A store running Stripe plus Shopify on Synder is plausibly looking at $65 to $275 a month for the connector, plus another $115 for QBO, just to do what most founders thought was one product.

Synder’s pricing page has the current numbers. Check them before you sign up, because they have crept up every year.

### QBO sets the ceiling on everything else

Inventory? Whatever QBO inventory does, which is not much, and not enough for anyone making a physical product. Multi-location stock? QBO. Aging reports? QBO. Custom GL rules? Whatever QBO lets you configure in its dropdowns. Synder is constrained by QBO’s data model, and so are you. The day you outgrow QBO, you do not just outgrow Synder. You outgrow the whole stack and have to migrate twice.

### Stripe Connect support is shallow

If you run a marketplace on Stripe Connect with platform fees, application fees, transfers, and destination charges, the standard Synder mapping struggles. The full Connect picture (platform fee revenue, Connect liability, payout to connected accounts, fee math per leg) is something most connector tools handle as an afterthought. ERPClaw handles it as a first-class part of the Stripe integration, because we built it for the SaaS and platform crowd from day one.

### Daily summary mode kills the audit trail

Synder offers per-transaction posting, which is the right default. The problem is that per-transaction floods QBO with thousands of journal entries a month, and QBO is not a good place for that volume. Most merchants end up flipping to summary mode to keep QBO usable, and now they have lost the per-transaction detail they were paying Synder for.

### Closed source, in their cloud

Your books live in their cloud. The mapping logic is a black box. If Synder shuts down, gets acquired, pivots, or doubles prices next year, your only option is a CSV export and a fresh start with the next vendor.

None of this means Synder is bad. It means the bill is bigger than it looks, and the platform under the bill is not yours.

## What ERPClaw does instead

ERPClaw is built on the opposite idea. The accounting tool, the inventory tool, the ERP, and the general ledger are all the same product. There is nothing to sync to, because there is nothing on the other side of the sync. The Stripe integration writes journal entries straight into your books. The Shopify integration does the same. Both share one chart of accounts. Both reconcile against the same bank deposits. No connector. No destination tool. One tool.

The numbers, briefly:

- Stripe integration: payment sync, refunds, disputes, payouts with three-layer reconciliation, Connect platform fees, application fees, revenue recognition for SaaS, MRR and ARR reports, dunning. The integration is its own module with its own posting engine.

- Shopify integration: orders, refunds, payouts, disputes, gift cards, cost of goods, inventory, GDPR webhooks, status mirror, daemon-driven incremental sync, OAuth pairing through our own connector. Same shape: a real module writing real ledger entries.

- Full ERP scope for the rest of the stack: inventory, manufacturing, HR, payroll, CRM, projects, assets, quality, procurement. All sharing the same database, all owned by you, all included.

Pricing is the same number for everyone, which is zero. No per-seat fee. No per-integration fee. No upsell tier. No enterprise sales calls. The pricing page is a short page, and that is on purpose.

Self-hosted means the database file sits on your disk, on your machine. Your books are your books. If we vanish tomorrow, your data does not go with us. You can export it, fork the repo, hire someone else to run it, or just keep using the version you have. The open source license is a real guarantee, not a marketing line.

## The “two bills vs one tool” math, in dollars

This is the part I want you to actually do on a calculator, because it is the single biggest reason most Synder customers should at least look at ERPClaw.

Synder’s job in your stack is to translate Stripe and Shopify into QBO. Both ends of that translation cost money. Synder costs money. QBO costs money. You need both for the workflow to work, and you cannot drop either one without breaking everything.

ERPClaw collapses the entire translation. The thing Synder is translating into is built into ERPClaw itself. There is no QBO, so there is no QBO bill. There is no connector, so there is no connector bill. The Stripe and Shopify integrations are first-class modules of the ERP, not third-party add-ons.

Here is the math on a typical Stripe plus Shopify store today:

- Synder plan covering Stripe plus Shopify: about $115 to $275 a month (as of June 2026), depending on tier and volume.

- QuickBooks Online Plus: $115 a month.

- Total: roughly $230 to $390 a month, every month.

Call it $260 a month on average. That is $3,120 a year. Over five years, with QBO’s annual price hikes, you are well past $16,000. For plumbing. For double-entry bookkeeping that has not changed in 530 years.

If you are running a large business with a finance team, fine, go pay NetSuite. There is a much bigger group of stores for whom that plumbing bill is just a tax on not knowing the alternative exists. ERPClaw is the alternative. One tool. One install. Zero dollars. Your own server.

## Side-by-side comparison

The same numbers live on the /compare/synder page, but here they are inline for the people who scan.

| Feature
| Synder
| ERPClaw

| Price
| $65 to $275 per month, plus QBO
| $0 forever

| Stripe + Shopify support
| Yes
| Yes, deeper integrations

| Destination accounting
| QuickBooks Online required
| Built in (ERPClaw is the ERP)

| Number of tools required
| Synder + QBO = 2
| ERPClaw alone = 1

| Per-transaction GL detail
| Yes (defaults vary by plan)
| Yes, always

| Self-hosted
| No
| Yes

| Open source
| Proprietary
| open source

| Inventory + manufacturing
| Limited via QBO
| Full ERP scope

| HR / payroll / CRM
| No
| Yes

| Revenue recognition (SaaS)
| Manual or premium tier
| Built in

| MRR / ARR reports
| QBO export only
| Built in

| Stripe Connect platform fees
| Limited
| First class

| Customizable GL rules
| Limited
| Fully configurable

| Three-layer payout reconciliation
| No
| Yes

| Architecture
| AI-decorated (chat features added on top)
| AI-native (assistant is the primary interface)

| Database
| Their cloud (proprietary)
| SQLite or PostgreSQL via PyPika, on your machine

Fourteen rows. Every row is a place where the “two bills vs one tool” choice tilts in your favor.

## How to switch from Synder to ERPClaw

A full migration guide is coming as a separate post (`migrating-from-synder-to-erpclaw`, on the calendar for Q4), but the short version is this:

- Install ERPClaw on your own machine. Five minutes from a fresh laptop. Full instructions on the docs site.

- Pair the Stripe integration with your Stripe account. ERPClaw pulls historical events back as far as Stripe will let you, so your books catch up automatically.

- Pair the Shopify integration. It is shipped and installable today (the docs cover setup), or use a Custom App token if you prefer to skip the connector entirely.

- Bring your chart of accounts across. ERPClaw ships with a sensible default chart of accounts. If you want to mirror what was in QBO, the accounts table accepts your existing codes directly.

- Run both side by side for a month if it makes you feel safer. ERPClaw only reads from Stripe and Shopify; it does not interfere with Synder. Compare the totals at the end of the month. If the ERPClaw books match what Synder posted to QBO, cancel Synder.

- Cancel QBO when you are ready. ERPClaw is your books now.

Total switching cost: a couple of weekends and a moment of nerve when you click cancel on the QBO subscription.

## FAQ

Is ERPClaw actually free, or is it free until I get hooked?

Actually free. open source license, source at github.com/avansaber/erpclaw. There is no paywalled tier of accounting features. ERPClaw Cloud Managed, our hosted offering, is a paid product (onboarding by request through /pricing/), but self-hosting the full ERP is free, and that is the recommended path for most stores.

Does ERPClaw replace QuickBooks?

Yes. That is the whole point. ERPClaw is a real double-entry general ledger with chart of accounts, journal entries, AR, AP, bank reconciliation, financial statements, and the draft and submit lifecycle any accountant would expect. It is not a connector that posts into QBO. It is the place the books actually live.

What about my accountant? They only know QBO.

Any accountant who can read a trial balance can read ERPClaw’s trial balance, because it is the same trial balance. We export to CSV, JSON, and standard accounting formats. The chart of accounts is configurable, so it can match whatever your accountant is used to seeing. The data underneath is double-entry, just like every other accounting system on the planet. It is a transition, but it is not a re-education.

Does ERPClaw handle Stripe Connect for marketplaces?

Yes, as a first-class feature. Application fees, platform fees, transfers, destination charges, connected account payouts, and the GL posting for each leg. This is one place where Synder plus QBO is genuinely weak and ERPClaw is genuinely strong. If you run a Connect platform, start here.

Is the Shopify integration on the App Store?

Not yet; the public App Store listing is pending. The integration itself is shipped and installable today: the feature tour shows what it does and the docs cover setup. The pairing flow connects Shopify to your own ERPClaw instance without Shopify ever holding your keys, and there is also a self-hosted Custom App token flow if you prefer that route.

What if I outgrow SQLite?

You switch the backend. ERPClaw is database-agnostic via PyPika, which means the same code runs on SQLite (the default for self-hosted installs) or PostgreSQL (the option for high-write-concurrency or multi-tenant deployments). The migration is a configuration change, not a rewrite. The full reasoning is in why SQLite is the default. Either way, it is your database, on your own hardware. You are never stuck with us.

## The honest closing line

If you love Synder, keep using Synder. If wide marketplace coverage on top of QBO is exactly what you need, and you are happy paying for it, their product works.

If you have ever opened the QBO bill and the Synder bill in the same week and added them up and felt slightly insulted that this combination is what passes for “small business accounting” in 2026, ERPClaw is for you.

One tool, not two. Free. Open source. On your own server. AI-native, not AI-decorated. Same Stripe and Shopify depth, often deeper. The whole rest of the ERP included, because the modules are already in the box.

See the full Synder comparison.
Browse the Stripe integration.
Browse the Shopify integration.
Read the docs and install.
See the pricing page.

Welcome to the new shape of small-business accounting.

---

## Three Way Matching in Procure to Pay: A Plain English Guide
URL: https://www.erpclaw.ai/blog/three-way-matching-procure-to-pay-explained/

# Three Way Matching in Procure to Pay: A Plain English Guide
Three way matching catches paying for goods you never received and overpaying invoices. How the PO, receipt, and invoice match works, in plain English.

A vendor sends you an invoice for $4,800. The bookkeeper queues it for payment. Two weeks later you find out the warehouse only received $3,200 of stuff. The vendor short-shipped a pallet, never told you, and invoiced for the full order. You already paid the difference. Now you are chasing a $1,600 credit memo from a vendor who is suddenly hard to reach.

This is the everyday reason three way matching exists. It is the boring control that catches the dollars you would otherwise lose to short shipments, double invoices, price disputes, and quietly inflated bills. Once you have it set up, the bookkeeper stops being a single-point check on every payable. The system catches the mismatches automatically and flags only the few invoices that need a human.

This post covers what three way matching actually is, the dollar math on a real example, the three documents involved, what gets blocked vs released, the variance handling that real businesses hit weekly, and how to run it without spreadsheets.

## What three way matching means in plain English

Three way matching is the practice of comparing three documents before you pay a vendor invoice:

- The purchase order (PO) you sent the vendor. This is what you ordered, the agreed price, and the agreed quantity.

- The goods receipt (sometimes called a receiving note or GRN). This is what the warehouse actually unloaded from the truck, on what date, in what condition.

- The vendor invoice the supplier sent you, asking for payment.

The system checks that all three documents agree on what was ordered, what was received, and what is being billed. If they match, the invoice is approved for payment with no human intervention. If they disagree, the invoice is flagged for review.

That is it. The whole concept fits in one paragraph. The reason it has a fancy name is that the controls around it (who can release a flagged invoice, what tolerances are allowed, what gets posted to the GL when there is a mismatch) are where real businesses lose or save real money.

## A real dollar example, end to end

Let me walk you through one purchase from PO to payment with all three documents.

Step 1: you send a PO. You order 100 units of Widget X from Acme Co at $48 per unit. PO total: $4,800. Expected delivery: April 15.

Step 2: goods arrive. The truck shows up on April 17. The warehouse receives 80 units. The other 20 are on backorder. Receiving clerk creates the goods receipt: 80 units of Widget X received, condition OK.

Step 3: vendor invoice arrives. A week later, Acme sends Invoice #5512 for $4,800. The invoice claims 100 units shipped at $48 each.

The three way match runs:

| Document
| Quantity
| Price
| Total

| PO
| 100
| $48
| $4,800

| Receipt
| 80
| n/a
| n/a

| Invoice
| 100
| $48
| $4,800

The PO and the invoice agree. The receipt does not. The system blocks payment and routes the invoice to AP for review.

The fix. AP calls Acme, confirms 20 units were not shipped, and asks for a corrected invoice for $3,840 (80 units at $48). Acme issues a credit memo or revised invoice. The new invoice matches the receipt. Three way match passes. Payment releases.

Without three way matching, the bookkeeper would have queued the original $4,800 invoice for payment, paid it, and discovered the short shipment a month later when somebody noticed the warehouse was 20 units short on Widget X. The $960 overpayment becomes a vendor credit chase that takes weeks to resolve. Some of those credits never get recovered.

This is the whole point of the control. Catch the mismatch before the wire goes out, not after.

## The three documents, in detail

Let’s break down what each document is, who creates it, and what fields the match looks at.

### The purchase order (PO)

Created by your purchasing team or auto-generated from a low-stock alert. It records:

- Vendor: who you are buying from

- Items and quantities: SKU, description, quantity ordered

- Agreed price per unit: what you negotiated

- Expected delivery date

- Total PO value

- Approval status: most companies require a manager to approve POs above a threshold

Once a PO is approved, it is the contract. The vendor agreed to ship what is on it at that price.

### The goods receipt

Created by the warehouse or receiving team when the goods arrive. It records:

- Date received

- Items and quantities actually received (often less than ordered if there is a short ship)

- Condition: any damage, wrong items, missing parts

- Reference to the PO so the system knows which order this fulfills

- Receiver name for audit trail

The goods receipt is what your auditor uses to confirm you actually got what you paid for. If it does not exist, you cannot prove the goods were ever delivered.

### The vendor invoice

Sent by the vendor, recorded by AP. It records:

- Invoice number (the vendor’s, not yours)

- Invoice date (drives payment terms)

- Items and quantities billed

- Price per unit billed

- Tax, shipping, discounts

- Total amount due

- Reference to the PO if the vendor included one (good vendors do)

AP enters the invoice into the system and the three way match runs automatically.

## What the match actually checks

The match compares specific fields across the three documents. Different systems implement this slightly differently, but the core checks are universal:

Quantity match. Quantity invoiced must equal quantity received. If you received 80 and the vendor billed 100, the match fails. If you received 100 and the vendor billed 80, the match also fails (you actually owe more than they billed, but you should not pay until they correct the invoice).

Price match. Price per unit invoiced must equal price per unit on the PO. If your PO said $48 and the invoice says $52, the match fails. Vendors raise prices quietly all the time. Three way matching catches it.

Item match. The SKU on the invoice must match the SKU on the PO and the receipt. If the vendor swapped the item without telling you, the match fails.

Vendor match. The vendor on the invoice must match the vendor on the PO. Sounds obvious. It is not. Plenty of fraud cases involve a fake invoice from a lookalike vendor name routed through someone who does not check.

Tolerance check. Most companies allow a small variance, say 2% on price or 1 unit on quantity, to handle normal supplier variance. Anything inside the tolerance auto-approves. Anything outside flags for review.

When all four checks pass within tolerance, the invoice is released for payment. When any check fails, the invoice is held and routed to a queue.

## The variances real businesses hit weekly

In a real business, a clean three way match is the exception, not the rule. Here are the variance types you will deal with constantly and how the system should handle each.

### Price variance

Vendor billed $52 per unit but the PO said $48. Maybe a price increase, maybe an honest mistake, maybe a clerical error in their system.

Handling. AP calls the vendor or pulls up the contract. If the price increase was approved, AP updates the PO price (usually requires a manager) and the match passes. If the price was wrong, the vendor issues a corrected invoice.

The dollar number is real. A silent price increase creeping across a meaningful share of your suppliers, undetected for a year, is a measurable hit to gross margin.

### Quantity variance, short ship

Ordered 100, received 80, billed 100. Most common variance type.

Handling. Block the invoice. Receive the remaining 20 when they arrive (creating a second goods receipt against the same PO). Once total received hits 100, the original invoice matches and releases. If the remaining 20 will not ship, the vendor issues a credit memo for $960.

### Quantity variance, over ship

Ordered 100, received 110, billed 100. Vendor accidentally shipped extra.

Handling. Decide whether to keep the extra (and pay for it, with an updated invoice from the vendor) or return it. Goods receipt records 110, AP holds the original invoice, AP either gets a new invoice for 110 or arranges a return.

### Item substitution

Ordered Widget X but the receipt says Widget Y. Vendor substituted because Widget X was out of stock.

Handling. Block. Confirm with purchasing whether the substitute was authorized. If yes, change the PO. If no, return.

### Freight or tax mismatch

PO totaled $4,800 in line items. Invoice is $4,920 because the vendor added freight or tax that was not on the PO.

Handling. This is where tolerance bands matter. Many companies set a freight tolerance of $200 or 5% of PO value to auto-pass freight charges that line up with the shipping terms.

### Duplicate invoice

Vendor accidentally sent the same invoice number twice. Or sent the same charges with two different invoice numbers.

Handling. Most systems block on invoice number duplication automatically. The version with two different numbers is the dangerous one and is exactly what duplicate detection tools catch.

## What the GL looks like in a three way match flow

The accounting moves through three GL postings, not one. This is the part most people skip past, and it is the part that matters most for clean books.

On goods receipt. The warehouse received goods you have not yet been billed for. You owe somebody for them, but you do not yet have an invoice. The system posts:

`DR Inventory $3,840
CR GR/IR (Goods Received Not Invoiced) $3,840`
GR/IR is a clearing account. It says “we owe somebody for these goods, vendor invoice pending.”

On invoice receipt and three way match pass. The invoice arrives, matches, releases. The system posts:

`DR GR/IR $3,840
CR Accounts Payable $3,840`
The GR/IR clearing account zeros out. The AP balance goes up by what you actually owe.

On payment. When you pay the invoice:

`DR Accounts Payable $3,840
CR Cash $3,840`
This three step flow gives you visibility into “what have we received but not yet been billed for” (the GR/IR balance), which is one of the most useful AP control numbers and one most small businesses never look at.

## Why every business above $2M revenue needs this

Below roughly $2M revenue, most companies do not formally run three way matching. The CEO knows every PO. The bookkeeper checks every invoice. The warehouse is the founder’s brother. Controls are by relationship, not by process.

Above $2M, the wheels come off. Multiple people place orders. The warehouse hires a part-timer. The bookkeeper gets behind. Vendors quietly raise prices. Short shipments stop getting flagged. The first time you do a real inventory count and find $40,000 less stock than the books say, you realize the controls failed three months ago.

Three way matching is the cheapest insurance policy against that scenario. It does not require a CFO. It does not require a $5,000 a month AP automation tool. It requires a system that creates POs, records goods receipts, and matches them to invoices. Most ERPs include this.

## How ERPClaw handles three way matching

ERPClaw is an open-source ERP with three way matching in the core. Here is the workflow:

Step 1: create the PO. Use the `add-purchase-order` action with vendor, items, quantities, prices. Manager approval required for POs over a configurable threshold.

Step 2: receive the goods. When the truck arrives, the warehouse runs `add-goods-receipt` referencing the PO. Records actual quantities received. The system posts the GR/IR journal entry automatically.

Step 3: enter the vendor invoice. AP runs `add-purchase-invoice` referencing the PO. The three way match runs automatically. If quantity, price, and item all match within tolerance, the invoice releases for payment. If any check fails, the invoice goes to a hold queue.

Step 4: review the hold queue. AP pulls up flagged invoices, reviews the variance, takes action (call the vendor, update the PO, issue a credit memo). Anything cleared releases for payment.

Step 5: pay. When invoices are due (per terms), the `submit-payment-batch` action generates payment files and posts the GL.

The whole flow uses the purchasing module plus the accounting module, both shipped in the core ERPClaw install. Free, open source, runs on SQLite or PostgreSQL.

## What you save by running this in software vs spreadsheets

A lot of small businesses still run procure-to-pay in spreadsheets. POs in a Google Sheet, receipts in another sheet, invoices in QuickBooks, matching done by eyeballing. This is the worst possible setup because nothing is enforced.

Real businesses lose money this way. The dollar amounts are small per incident (a $200 short ship, a $50 silent price increase) but they happen weekly. Across a year, a mid-sized company running this manually loses a real slice of its cost of goods to undetected variances. That is $20,000 to $60,000 a year in real money, gone, because the controls were not enforced.

Three way matching in software is one of the highest-ROI setups a small business can make. The implementation cost in ERPClaw is the time it takes to enter your existing vendors and one PO. The savings start the same week.

## CTA

If you want three way matching in your business without paying QuickBooks Online Advanced ($235/mo) or NetSuite ($25,000+/yr) for it, ERPClaw includes it free in the core install.

- Install ERPClaw (5 minutes)

- Try the demo (browser, no install)

- See the QuickBooks alternative comparison

- Read about AI-native ERP architecture

- See pricing (it is $0)

ERPClaw is open source, AI-native, and ships with the procurement, inventory, and accounting modules in one shared database. No per-user fee, no upgrade tier behind a paywall, no enterprise gate.

## FAQ

### Is three way matching the same as four way matching?

Four way matching adds a fourth document, the inspection or quality check, to the three way match. You match PO + receipt + inspection + invoice. It is used in industries where goods can be received but later fail QC (regulated manufacturing, pharmaceuticals, food). Three way matching is enough for most small to mid-market businesses.

### Do I need three way matching for service vendors?

Service invoices (consultants, software subscriptions, marketing agencies) usually do not have a goods receipt because nothing physical was delivered. For services, most systems use a two way match (PO + invoice) or a contract-based match (compare invoice to the SOW or recurring contract). Three way matching is for physical goods.

### What is GR/IR and why does it matter?

GR/IR stands for “Goods Received but not Invoiced.” It is a clearing account that holds the value of goods you received before the vendor invoice arrives. The balance tells you “we have inventory we have not yet been billed for, dollar value $X.” It is one of the most useful control numbers in AP and one most spreadsheet-run businesses never see.

### How tight should my tolerances be?

Common defaults are a small tolerance band on price, a unit or so on quantity, and a somewhat wider band on freight. Tighter tolerances catch more variances but flag more invoices for human review. Looser tolerances release more invoices automatically but let small overcharges through. Most companies tune the tolerance per vendor based on supplier reliability.

### Does ERPClaw three way matching work with PostgreSQL?

Yes. ERPClaw is database-agnostic via PyPika. SQLite is the default for install simplicity, but PostgreSQL is fully supported as an alternative backend. The same procurement actions run on either database.

### Can I run three way matching without a formal warehouse?

Yes. The “warehouse” can be one person on the receiving dock, the office manager, or even the founder. The point is that whoever physically takes possession of the goods enters the goods receipt before AP processes the invoice. Even a one-person operation benefits because it forces a check before payment.

### Does ERPClaw support backorders and partial receipts?

Yes. A PO can have multiple goods receipts against it. If you order 100 units and receive 80 today, 20 next week, you create two receipts. The system tracks remaining quantity automatically and the three way match runs against the cumulative received total.

Install ERPClaw and try a real three way match in your own data.

---

## Webgility Alternative: Free Shopify ERP, QBO Included
URL: https://www.erpclaw.ai/blog/webgility-alternative-free-shopify-erp/

# Webgility Alternative: Free Shopify ERP, QBO Included
Webgility runs $59 to $499 a month to sync Shopify into QuickBooks. ERPClaw replaces both for $0: the sync and the books, in one system.

If you sell on Shopify and Amazon and eBay, you probably already know what Webgility does. It pulls every order from every channel, formats it the way QuickBooks wants it, and pushes it across so your books actually balance. For multi-channel sellers, it has been the default answer for years.

It is also two bills. Webgility runs $39 to $249 a month depending on volume. QuickBooks Online sits underneath it at another $115 a month. Add Bill.com or a payroll add-on and you are over $300 a month before you have sold a thing. That is the part nobody mentions in the sales call.

This post is the long-form Webgility alternative pitch from someone who built one. ERPClaw is free, open source, AI-native, and its Shopify integration is shipped and installable today. For Shopify-primary sellers, it does the multi-channel sync job AND replaces QuickBooks at the same time. One tool, one install, zero monthly bill. I will be honest about where Webgility still wins (deep Amazon and eBay support) so you can decide if the trade is worth it for your stack.

Webgility is competent at what it does, but it is AI-decorated software (a desktop-era connector with chat features added in the last 18 months). ERPClaw is the only AI-native option in this category, which means the AI assistant is the primary interface, the spec drives the code, and the architecture was built around AI from day one. The full argument is in AI-decorated vs AI-native software.

The shorter version of this comparison lives at the Webgility vs ERPClaw page. Keep reading for the full Webgility vs ERPClaw breakdown, real dollar math, and a migration walkthrough.

## What Webgility actually does well

A fair Webgility alternative post should start with what Webgility is good at, because they have earned a real customer base.

Channel breadth. Webgility connects Shopify, Amazon, eBay, Etsy, Walmart, BigCommerce, WooCommerce, Square, and others. If you sell on five rails, they have mapped every quirk by now. Amazon FBA fees, settlement reports, eBay managed payments, Etsy refunds. The kind of edge cases a new tool would get wrong on day one.

Inventory sync across channels. Sell the same SKU on Shopify and Amazon? Webgility keeps the stock count consistent so you do not oversell on one channel while another shows it as available. Real workflow, real value.

QuickBooks mapping depth. Years of QBO posting logic baked in. Sales tax by jurisdiction, Amazon’s odd settlement format, Shopify payouts. The kind of thing where if you tried it in spreadsheets you would lose a week per channel.

If you are running a serious multi-channel operation across Amazon and eBay and want everything in QBO, Webgility does that job well. The question is whether the bill matches the value, and whether QBO is actually the right place for your books to live.

## The QBO bottleneck nobody talks about

Webgility’s whole product depends on QuickBooks Online sitting on the other end. That is not a side note. It is the shape of the company. Without QBO, Webgility has nowhere to put your data. The connector has nothing to connect to.

That dependency creates four problems you may not have thought through.

You pay twice, every month, forever. Webgility’s plans run from $39 to $249 a month. QBO Plus is $115 a month (as of June 2026) and rises every year. Multi-channel sellers usually land on the $99 or $179 Webgility tier, so you are looking at $200 to $350 a month total just for the connector and the destination. Five years of that is $12,000 to $21,000 in software costs for what most founders thought was one product.

QBO sets the ceiling on everything else. Inventory? Whatever QBO Inventory does, which struggles past a few hundred SKUs. Multi-warehouse? QBO does not really do it. Custom GL rules? Whatever QBO lets you click in a dropdown. Webgility cannot send anything QBO cannot accept. The day you outgrow QBO, you outgrow the whole stack and have to rip it all out.

Per-transaction sync floods QBO. 5,000 orders a month means 15,000+ journal entries hitting QBO, which is not built for that volume. Most sellers flip to summary mode and lose the per-order detail they were paying for.

Two black boxes stacked. Both Webgility and QBO are closed source. When the totals do not tie out at month-end, you cannot tell which side dropped the transaction. You file a ticket and wait.

## What ERPClaw replaces

ERPClaw is built on the opposite assumption. The connector and the accounting tool and the inventory module are all the same product. There is nothing to sync to, because there is nothing on the other side of the sync. The Shopify integration writes journal entries straight into your books. The chart of accounts, the trial balance, the inventory ledger, the payout reconciliation. All of it lives in one database on your own machine (SQLite by default, or PostgreSQL if you prefer).

Concretely, switching to ERPClaw from a Webgility plus QBO stack collapses three line items into one tool:

- Webgility (gone). Shopify orders post directly to your ledger via the Shopify integration, shipped and installable today. All core Shopify actions, OAuth pairing through the pairing-code bridge, daemon-driven incremental sync, GDPR webhooks, three-layer payout reconciliation.

- QuickBooks Online (gone). ERPClaw IS the books. Real double-entry, immutable submitted entries, draft and submit lifecycle, AR, AP, bank rec, P&L, balance sheet, cash flow. Same numbers your accountant already knows how to read.

- Standalone inventory module (gone). Built-in multi-warehouse inventory with FIFO valuation, item-level cost tracking, low-stock alerts, transfer orders. No more wedging inventory into QBO’s data model.

Stripe is a first-class part of the same picture. If you also take payments through Stripe (plenty of stores run a separate Stripe for subscriptions or marketplace flows alongside Shopify Payments), the Stripe integration ships with all core Stripe actions including ASC 606 revenue recognition, Connect platform fees, and dunning.

Pricing is the same number for everyone, which is zero. No per-channel fee. No per-seat fee. No premium tier for revenue recognition. The pricing page is a short page on purpose.

## Where Webgility still wins (the honest part)

I am not going to pretend ERPClaw has feature parity with Webgility on every channel. Here is where Webgility is still the better answer in v1:

Deep Amazon support. Webgility handles Amazon FBA fees, settlement reports, FBM, multi-channel fulfillment, and various Amazon marketplace quirks (US, CA, UK, EU) with mappings refined over a decade. ERPClaw v1 does not have a native Amazon connector yet. If Amazon is more than a side channel for you, this matters.

Native eBay, Etsy, Walmart, BigCommerce sync. Webgility maps managed payments, dispute flows, promoted listings fees, marketplace-specific deductions. ERPClaw v1 ships with Shopify only.

Cross-channel inventory rebalancing. If you need stock counts to update on Amazon the second a Shopify order lands, Webgility does that today. ERPClaw holds the stock in one place but does not push counts back out to Amazon and eBay yet.

So the honest version: if your business is mostly Shopify and the rest is small (some Amazon you handle by import, an eBay store that is not your main hustle), ERPClaw replaces the whole stack and saves you $200 to $350 a month. If you are genuinely split across four marketplaces with Amazon at 40% of revenue, stay on Webgility for now; multi-marketplace breadth is genuinely theirs, and I am not going to oversell what we have today.

## Side-by-side comparison

The same numbers live on the Webgility comparison page. Here they are inline for the people who scan.

| Feature
| Webgility
| ERPClaw

| Price
| $39 to $249 per month
| $0 forever

| Plus QuickBooks subscription
| Required ($115/mo+)
| Not needed

| Sales channels
| Shopify, Amazon, eBay, Etsy, etc.
| Shopify in v1

| Per-transaction sync
| Yes
| Yes

| COGS tracking
| Yes
| Yes (multi-warehouse)

| Inventory module
| QuickBooks-dependent
| Built-in

| Self-hosted
| No (Webgility cloud)
| Yes

| Open source
| Proprietary
| open source

| Architecture
| AI-decorated (chat features added on top)
| AI-native (assistant is the primary interface)

| Database
| Their cloud
| SQLite or PostgreSQL via PyPika, on your machine

| Two tools or one
| Two (Webgility + QBO)
| One (ERPClaw)

| Multi-channel future
| Strong (many integrations)
| Coming (Stripe shipped, more in roadmap)

Ten rows. The one ERPClaw loses on today is channel breadth. The other nine all tilt the other way.

## The dollar math, on a calculator

This is the part I want you to actually do, because it is the single biggest reason most Webgility customers should at least look at the alternative.

Say a typical mid-volume multi-channel store does $100k a month in revenue, mostly Shopify with some Amazon and eBay on the side.

- Webgility Pro tier: about $179 a month.

- QuickBooks Online Plus: $115 a month, rising every year.

- Total: $294 a month, every month, before you sell a thing.

That is $3,528 a year. Over five years with QBO’s annual price hikes (they have raised QBO Plus three times since 2022), you are well past $18,000. For plumbing. For double-entry bookkeeping that has not changed in 530 years.

Now run the same store on ERPClaw plus Stripe (if you have it):

- ERPClaw: $0.

- Stripe integration: $0.

- Shopify integration: $0.

- The whole ERP, every module: $0.

- Total: $0 a month, every month, forever.

The trade-off, again, is channel breadth in v1. If Shopify is the bulk of your volume, you keep $294 a month. If Amazon is genuinely 50% of your revenue, the trade is harder and you should probably stay on Webgility for now.

You can check Webgility’s current pricing on their pricing page before deciding. They have crept it up a few times in the last two years.

## When to stay on Webgility

I am writing a Webgility alternative post but I am not going to pretend the answer is always “switch.” Here is when you should stay:

- Amazon is a large share of your revenue. ERPClaw v1 does not have a native Amazon connector. You would have to import settlements manually or wait for the connector to ship.

- You actively rebalance inventory between channels. Webgility’s cross-channel stock sync is a real workflow that ERPClaw does not match yet.

- Etsy, Walmart, or BigCommerce is your main store. Same story. We start with Shopify; the others are roadmap.

- You have a bookkeeper who lives in QBO and refuses to move. The migration is not free. If your accountant fights you, the friction may not be worth $200 a month in your case.

- You need the QBO ecosystem of third-party apps. A lot of small business apps integrate with QBO and not with anything else. ERPClaw has its own module library, but if you depend on a specific QBO app, factor that in.

If none of those apply, the Webgility plus QBO bill is mostly money you are paying for nothing your business actually needs.

## How to switch from Webgility to ERPClaw

A full migration guide is on the calendar for Q4. The short version:

- Install ERPClaw on your own machine. Five minutes from a fresh laptop. Full instructions in the Shopify install walkthrough.

- Pair the Shopify integration. It is shipped and installable today; the feature tour shows what it does and the docs cover setup. The pairing flow connects Shopify to your own ERPClaw instance without Shopify ever holding your keys, and there is also a self-hosted Custom App token flow if you prefer that route.

- Bring your chart of accounts across. ERPClaw ships with a sensible default chart of accounts and accepts your existing QBO codes directly if you want to mirror them.

- Import historical orders. ERPClaw pulls Shopify orders back as far as the API allows. For Amazon and eBay history you would need a one-time CSV import via the generic importer.

- Run both side by side for a month. ERPClaw only reads from Shopify, so it does not interfere with Webgility. If the totals match at month-end, you are clean.

- Cancel Webgility, then cancel QBO. In that order. ERPClaw is your books now.

Total switching cost: a couple of weekends and a moment of nerve when you click cancel on QBO.

## FAQ

Is ERPClaw really a Webgility alternative if it does not support Amazon and eBay yet?

For Shopify-primary sellers, yes. For sellers where Amazon or eBay is the main channel, Webgility’s multi-marketplace coverage is genuinely broader, and I would rather tell you that upfront than oversell. The Stripe and Shopify integrations are deep and shipping today.

Does ERPClaw really replace QuickBooks?

Yes. That is the whole point of “Webgility alternative” being framed this way. ERPClaw is a real double-entry general ledger with chart of accounts, journal entries, AR, AP, bank reconciliation, financial statements, and the draft and submit lifecycle any accountant would expect. It is not a connector that posts into QBO. It is the place the books actually live.

What about my accountant? They only know QBO.

Any accountant who can read a trial balance can read ERPClaw’s trial balance, because it is the same trial balance. We export to CSV, JSON, and standard accounting formats. The chart of accounts is configurable, so it can match whatever your accountant is used to seeing. The data underneath is double-entry, just like every accounting system on the planet. It is a transition, not a re-education.

How does ERPClaw compare to Synder, the other big QuickBooks ecommerce sync alternative?

Synder targets a different cross-section (Stripe plus Shopify into QBO, mostly for SaaS and DTC) and Webgility leans more multi-channel marketplace. We have a separate post on the Synder vs ERPClaw comparison if you are evaluating both.

Is the Shopify integration on the App Store?

Not yet; the public App Store listing is pending. The integration itself is shipped and installable today: the feature tour shows what it does and the docs cover setup. The pairing flow connects Shopify to your own ERPClaw instance without Shopify ever holding your keys, and there is also a self-hosted Custom App token flow if you prefer that route.

What if I outgrow SQLite at high volume?

You switch the backend to PostgreSQL. ERPClaw is database-agnostic via PyPika, so the same code runs on either engine. SQLite is the default because it fits the workload of a self-hosted store; PostgreSQL is fully supported for the deployments that need it. The migration is a configuration change, not a rewrite. The full reasoning is in why SQLite is the default. Either way, it is your database on your own hardware. You are never stuck with us.

## The honest closing line

If you sell heavily across four marketplaces and Webgility plus QBO is doing the job, keep paying for it. The product works.

If you are mostly a Shopify store with some side-channel volume, and you have ever added up the Webgility bill and the QBO bill and felt slightly insulted that this is what passes for small business accounting in 2026, ERPClaw is for you.

One tool, not two. Free. Open source. AI-native rather than AI-decorated. On your own server. Same Shopify depth, full ERP included.

See the Webgility comparison page.
Browse the Shopify integration.
Read the install walkthrough.
Compare against Synder too.
See the pricing page.

Welcome to the new shape of multi-channel accounting.

---

## When Xero and QuickBooks Stop Fitting (and What Comes Next)
URL: https://www.erpclaw.ai/blog/when-xero-and-quickbooks-stop-fitting/

# When Xero and QuickBooks Stop Fitting (and What Comes Next)
Xero and QuickBooks work well for most small businesses until they don't. Five growth triggers that signal you've outgrown them, and the honest alternatives in 2026.

Xero and QuickBooks are the two products most small businesses run their books on, and for good reason. They handle the common cases (categorization, invoicing, bank feeds, payroll integration, year-end tax export) well enough that for most businesses, most of the time, there is no reason to look further.

This post is for the cases where there is. Five growth triggers where the products start to feel like they are working against you, and the honest options for what comes next.

## What Xero and QuickBooks do well

Before the criticism, the credit. Both products earned their position. The things they do well are exactly what an SMB in years 0-5 needs.

Bank feed automation. Both connect to most major banks; transactions land categorized and ready to confirm. The friction in the bookkeeping cycle is low; the categorization rules learn over time.

Invoicing. Both produce clean invoices with payment links, AR aging, and reminder automation. A small business that uses the invoicing properly meaningfully accelerates collections.

Basic financial reporting. P&L, balance sheet, cash flow statement, all generated automatically and usually correctly. The reports are good enough for the owner to make weekly decisions and for the accountant to file taxes at year end.

Payroll integration. Through their own tools (QuickBooks Payroll) or through partners (Gusto, ADP, RazorpayX, KeyPay). The integration is solid; payroll feeds the books cleanly.

Ecosystem. Hundreds of apps integrate. Whatever niche functionality you need (e-commerce sync, time tracking, expense capture, project profitability), there is an app that bolts on.

For a 2-30 person business in services or simple retail, this is the right tooling. Looking elsewhere wastes time.

## Five growth triggers that signal you have outgrown them

### 1. Multi-entity consolidation

Once you operate more than one legal entity (parent company plus subsidiaries, multiple geographic operations, separate funding vehicles for different business lines), consolidating financials in Xero or QuickBooks is exporting CSVs and stitching them in Excel. The native consolidation features are thin to nonexistent.

The hack works for two to three entities for a year or two. The hack stops working when the entities have intercompany transactions, when the auditor wants journal entries traceable across entities, or when the CFO wants real-time consolidated reporting instead of monthly Excel rollups.

Specific signal: you have an Excel file called “consolidation” that someone updates manually each month and it takes more than half a day.

### 2. Inventory and cost accounting at scale

QuickBooks has inventory features; Xero has them through add-ons. Both work for a product business with one location, a few hundred SKUs, FIFO or weighted-average costing, and standard sales channels.

Both start straining around 1,000-3,000 SKUs, multi-location inventory, kitting and bundling, landed cost allocation, or any cost accounting method beyond the defaults. The accountant patches with adjusting journals. The accounting becomes harder to audit because the patches accumulate.

Specific signal: month-end close requires inventory adjustments that take more than a day, or the inventory subledger and the GL disagree by amounts you cannot easily explain.

### 3. AI-strategic accounting workflows

If your business has structured opportunities for AI in the accounting workflow (transaction categorization at high volume, anomaly detection, automated reconciliation, narrative draft generation), Xero and QuickBooks have added AI features but they are bolt-ons. The architecture under them was designed in 1998 and the AI sits on top.

This shows up as: the AI features work for simple cases but get unreliable for the cases that would actually save time; the AI cannot reach into the underlying ledger structure deeply enough to drive workflows end-to-end; you are essentially using the AI as a smart suggestion engine that an accountant still re-does manually.

Specific signal: you evaluated the AI features, found them useful for 5-10 minutes per week, and decided they did not move the needle.

### 4. Heavy automation requirements

When your accounting needs to integrate deeply with custom workflows (a specific approval routing for AP, a tight integration with an industry-specific operations tool, a non-standard revenue recognition schedule), Xero and QuickBooks force you into the app marketplace. The marketplace apps work but the integration layer is brittle. Webhooks fail; data syncs lag; reconciliation requires manual fixes.

Specific signal: you maintain a folder of process documentation for “what to do when [integration] breaks,” and someone refers to it more than once a quarter.

### 5. Industry-specific or regulatory complexity

Specific industries (medical practices with insurance billing, construction with progress billing and percentage-of-completion accounting, SaaS with ASC 606 revenue recognition, multi-jurisdictional businesses with complex sales tax) push past what Xero or QuickBooks handle natively. Workarounds exist; their cost (in time, audit risk, and bookkeeper hours) grows with revenue.

Specific signal: your accountant repeatedly mentions that “we should really be using something different at your stage” but you have not figured out what.

## The honest options when you outgrow them

When one or more of the above triggers fire, the alternatives split into three categories.

### Move up the SaaS stack

Sage Intacct. Mid-market financial system. Native multi-entity consolidation. Strong audit-grade controls. Deep integration with the common middleware and ERP stack. Pricing is quote-based, sits well above QuickBooks/Xero tiers, and scales with entities and users. Standard recommendation for mid-market businesses outgrowing QuickBooks/Xero.

NetSuite. Oracle’s mid-to-large-market ERP. Broader scope than Sage Intacct (full ERP with inventory, manufacturing, CRM). More expensive. The recommendation for businesses that need ERP, not just accounting.

Rillet, DualEntry, Campfire. Newer AI-native finance close tools for SaaS and tech-forward businesses. Strong AI in the close cycle, less mature in inventory or industry-specific workflows. Good fit for businesses that want AI-native but stay in the SaaS-product business shape.

### Move to ERP

ERPClaw. AvanSaber’s open-source AI-native ERP. Built for businesses where accounting is integrated with operations (inventory, fulfillment, customer-facing surfaces), and where AI is the primary user pattern not a chat sidebar. Open source, no SaaS licensing fees. Good fit for businesses that have outgrown QuickBooks/Xero on multiple dimensions at once.

Doss. Agent-orchestrated workflow ERP. Strong on the workflow side; good for businesses where the bottleneck is process automation more than accounting depth.

Odoo. Established open-source ERP. Broader scope than ERPClaw but architecturally older; the AI features are layered on rather than native. Fit for businesses that want self-hosted ERP without strong AI requirements.

### Move sideways to specialized vertical tools

If your trigger is purely industry-specific (medical billing, construction accounting, agency time-and-billing), there are vertical SaaS products built around your specific workflow that include their own accounting. Sometimes the right answer is the vertical tool plus a thin general-ledger system rather than a horizontal ERP plus industry modules.

## How to evaluate the migration

Three questions cut through the demo theater.

What is your specific trigger? Map your friction back to one of the five categories above. If you cannot, you may not have outgrown your current stack; you may have a process problem masquerading as a software problem.

What does the migration actually look like in your case? Specifically, the chart of accounts mapping, the historical data import, the cutover timing, and the training. Vendors quote installation in weeks; reality for a typical SMB migration is 3-6 months end-to-end. Budget accordingly.

What is the total cost of staying? This is the question owners often skip. The Excel consolidation that takes half a day each month, the inventory adjustments that take a day at close, the brittle integrations that fail quarterly. Quantify the hours; multiply by what an FTE costs. The comparison to the migration cost is usually closer than it feels.

## Where ERPClaw fits

ERPClaw is the right next step when triggers 1, 2, 3, and 4 fire together. The business has outgrown horizontal accounting tools, needs ERP capability, wants AI-native rather than AI-decorated, and prefers open source over SaaS lock-in.

ERPClaw is not the right step when only trigger 5 fires (industry-specific complexity in a domain ERPClaw does not specifically serve), when the business is small enough that mid-market ERP overhead exceeds the benefit, or when the team’s preference is for fully-managed SaaS over self-hosted open source.

The honest comparison is in the 5 AI-native ERPs that earn the label post: where ERPClaw genuinely beats the alternatives, and where the alternatives beat it. The product that wins on all dimensions does not exist. The product that wins on your dimensions might.

## Closing

Xero and QuickBooks are fine. Until they are not. The five triggers above are the signals to look for. The alternatives are real and they have genuinely different shapes. The migration is non-trivial and the decision should be conscious.

If you are evaluating, the framework here is yours to use. If you want help executing, the AvanSaber and ERPClaw teams are reachable through the portfolio.

Tagsxero-alternativequickbooks-alternativeai-nativeerpmigration
Share:XLinkedInHacker News

---

## Where AI Belongs in Accounting (And Where It Doesn't)
URL: https://www.erpclaw.ai/blog/where-ai-belongs-in-accounting/

# Where AI Belongs in Accounting (And Where It Doesn't)
Three places AI earns its keep in accounting and three places it doesn't. A practitioner framework with the decision tree, not the marketing pitch.

Every accounting software vendor in 2026 markets AI features. The honest version is that some of those features earn their place and some are a chat sidebar bolted onto a 1998 product. This post is the framework we use at ERPClaw to decide which workflows AI actually improves and which it makes worse.

The framework has six categories: three where AI earns its keep, three where it does not. The categories are not about AI capability. They are about the structure of the accounting task and the consequences of getting it wrong.

## Three places AI earns its keep

### 1. Transaction categorization at scale

A small business produces hundreds of bank transactions per month. A growing business produces thousands. Each one needs a category: software subscription, vendor payment, payroll, owner draw, refund.

The traditional approach is a rules engine. Match by vendor name, then by amount band, then by memo text. Rules engines plateau because vendor names drift, memo text varies, and new vendors arrive constantly.

A model trained on the business’s own categorization history keeps improving as corrections feed back in. The category space is finite and well-defined. The decision is reversible (re-categorize). The cost of an individual wrong categorization is low. The volume is high enough that human review of every transaction is uneconomic.

This is the canonical AI-belongs case. It is also the case most “AI accounting software” actually delivers on, because it is the easiest to evaluate against ground truth.

### 2. Anomaly detection

A bookkeeper looking at 4,000 transactions per month is not going to notice that one of them is structurally weird. A model can.

Weird looks like: a vendor name that has never appeared before paying an amount in the range of legitimate vendors; a transaction that fits the pattern of an internal control violation (round numbers, even amounts, just below an approval threshold); a sequence of transactions to a new vendor that escalate in amount over weeks.

The model does not decide that the transaction is fraud. It flags it for human attention. This is Filter mode (in the enterprise decisioning sense): the model separates the work that needs human attention from the work that does not. The human still investigates.

What earns AI its keep here is that the alternative is not no detection. It is detection at audit time, three months later, when the fraud has already happened.

### 3. Draft generation for narrative artifacts

Accounting produces narrative artifacts: month-end commentary, variance explanations, executive summaries, audit responses. A controller writes these on top of structured ledger data. The structured data is the input. The narrative is the output.

A model that reads the ledger and drafts the narrative saves an hour of work per artifact. The controller edits the draft, adds context the model cannot see, and signs. The model never signs the artifact and never sends it.

The value here is not the writing. The value is the time the controller saves. The controller still owns the artifact’s correctness. The model just reduces the time from blank page to first draft.

## Three places AI doesn’t belong

### 1. Anything that signs

Tax returns get signed by a preparer. Audit reports get signed by an auditor. Financial statements get signed by a controller or CFO. Sales tax filings get signed by an officer of the company.

The signature is not a UX detail. It is the legal mechanism by which a human person takes responsibility for the document’s accuracy. The IRS does not accept a model’s signature. The state department of revenue does not accept a model’s signature. A bank evaluating a loan application does not accept a model’s signature.

A model can draft any of these. A model can pre-fill any of these. A model cannot sign any of these. Software vendors that imply otherwise are setting their customers up for a problem at filing season.

### 2. Source-of-truth reconciliation

Reconciliation answers the question: do our books match the bank, the credit card processor, the merchant of record, the prior-period ledger? When the answer is “no,” reconciliation also answers: what is the difference, and what should we do about it.

The “what is the difference” half can be partly automated. A model can list the unmatched items. But the “what should we do about it” half requires judgment about which side has the right number. The bank might be wrong (processing delay, batch error, missing transaction). The books might be wrong (timing, classification, duplicate entry). The merchant of record might have a different fee structure than expected.

Resolving this requires reading documents, calling people, asking questions, knowing the business. A model can support the work; it cannot do the work. Vendors that claim “automated reconciliation” are usually doing the easy half (matching) and either skipping or hand-waving the hard half (resolution).

### 3. Decisions that depend on what the business does next

Revenue recognition depends on what the customer contract says and what the company commits to deliver. Lease accounting depends on what the terms of the lease are and what the company’s intent is. Inventory valuation method depends on what the business has committed to in past statements.

These are not data-extraction problems. They are interpretation problems. The model can read the contract; it cannot decide what the business will do under the contract. The model can read the lease; it cannot decide whether the company will exercise the renewal option. The model can compute LIFO or FIFO or weighted-average; it cannot decide which method is appropriate.

The accountant’s value here is interpretation. The model is, at best, a faster reader.

## The decision tree we use

Three questions, in order.

- Is the workflow’s output signed by a human in a legal or regulatory sense? If yes, the model can prepare and draft. It cannot decide. Final decisions stay with the human signer.

- Does the workflow’s correct answer depend on facts outside the ledger (contracts, communications, intent, future plans)? If yes, the model can extract candidates but cannot resolve them. A human reconciles.

- Is the workflow volume high enough that human review of every instance is uneconomic? If yes, AI in Filter or Recommender mode (with a risk-scored review queue) is the right shape. If no, AI is at best a draft tool and may not earn its keep at all.

The interesting cases are the ones that pass question 3 but fail 1 or 2. High-volume signed work (sales tax filings) belongs at draft-and-sign, never at auto-file. High-volume interpretation-dependent work (revenue recognition at scale) belongs at recommend-and-confirm, never at auto-book.

## ERPClaw’s stance

ERPClaw is built on this framework. The AI handles categorization, anomaly detection, draft generation, and the support work around reconciliation. The AI does not sign tax returns, does not auto-resolve recon breaks, and does not pick revenue recognition method. The signature, the resolution, and the method choice live with the accountant who is licensed to make them.

This is what we mean when we say “AI-native that earns the label.” The AI is not a chat sidebar. It is the way the workflow is structured. But it stops at the boundary where a human signature, a judgment call, or an interpretation lives.

The vendors that blur this boundary do their customers a disservice. The vendors that respect it produce software that an accountant can actually rely on.

If you are evaluating AI accounting software, ask the vendor where their AI signs versus drafts. The honest ones have an answer. The marketing-led ones do not.

Tagsai-nativeai-accountingframeworkvision
Share:XLinkedInHacker News

---

## Open Source Xero Alternative: Own Your Books, Pay $0
URL: https://www.erpclaw.ai/blog/xero-alternative-open-source/

# Open Source Xero Alternative: Own Your Books, Pay $0
Xero is great cloud accounting. Add Payroll, Projects, Expenses, and an inventory plugin and you are at $300+/mo. The open-source ERP that bundles it free.

Xero is good software. Let me say that up front, because most “Xero alternative” posts open with a list of complaints and that is not the honest take. If you are running a small business in Australia, New Zealand, or the UK, Xero is probably what your accountant set you up on, and they had a good reason. Bank feeds work, BAS and VAT returns work, the interface is pleasant, and there is a Xero-trained bookkeeper in every suburb. For a freelancer just needing clean books, Xero Early at $25 a month is fine.

This post is not for that person.

This post is for the person who started on Xero two or three years ago, kept growing, and is now staring at a stack of add-ons wondering when the bill got this high. You added Xero Payroll for your third hire. You added Xero Projects to bill clients by hour. You added Xero Expenses. You bought an inventory plugin from the App Marketplace because Xero itself does not really do inventory. You signed up for a CRM somewhere else. You are now paying $55 to $90 a month to Xero, plus $20 to $80 for the inventory plugin, plus a CRM bill, plus a bookkeeper to glue it all together. None of it is wasteful. All of it crept up one line at a time.

That is the moment you outgrew Xero. The accounting still works. The thing around the accounting is the problem. We wrote up how to recognize that moment, on either platform, in when Xero and QuickBooks stop fitting.

ERPClaw is what comes next as a Xero alternative open source option. It is free, open source, self-hosted, AI-native, and it covers accounting plus inventory plus manufacturing plus HR plus payroll plus CRM plus projects in one system with one database. The pitch is not “we do accounting better than Xero.” The pitch is “we do accounting and the eight other things you bolted on Xero with, all in one place, for zero per-user fees, forever, operated through your AI assistant rather than a 2010s forms UI.”

Xero is excellent at what it does, but it is AI-decorated software (a 2010 cloud accounting product with chat features added in the last two years). ERPClaw is the only AI-native option in this category, which means the architecture, the spec-first build, and the pricing all reflect a cost base Xero cannot match without rewriting from scratch. The full argument is in AI-decorated vs AI-native software.

## What Xero genuinely does well

Xero earned its reputation. To be specific:

Best cloud accounting product outside the US. In Australia, New Zealand, and the UK, bank feed coverage is excellent, BAS/VAT/MTD integrations are mature, and the local accountant network is enormous. Ask an accountant in Sydney or Manchester what to use, they will say Xero before you finish the question.

The interface is actually pleasant. Bank reconciliation is a nice flow. The mobile app works. Reports look like reports.

Multi-currency on the higher tier is solid. Invoice in three currencies with automatic gain/loss tracking on the Premium plan.

The accountant network in the Commonwealth is enormous. Tell your accountant “I’m on Xero” and they say “great, send me the advisor invite.” Tell them “I’m on Odoo” and they say “hmm, can you export a CSV.”

So the case for ERPClaw is not that Xero is broken. The case is that the Xero footprint is smaller than your business is the moment you have any of: physical inventory, manufacturing, project-based billing across more than a handful of jobs, more than 10 employees, multiple legal entities, or a CRM that needs to talk to invoicing.

## The Xero add-on pile

Here is the part the Xero pricing page does not spell out. Xero is a recurring monthly bill that scales with features, add-ons, and Marketplace plugins.

Current US pricing on xero.com/us/pricing (as of June 2026):

- Early: $25/month, 20 invoices, 5 bills

- Growing: $55/month, unlimited invoices and bills

- Established: $90/month, multi-currency and projects

The headline. Now look at the real monthly bill for a 15-person small business:

- Xero Established, $90

- Xero Payroll for 8 employees, $40 base plus $6/person = $88

- Xero Projects for 3 users, $7/user = $21

- Xero Expenses for 5 users, $4/user = $20

- Inventory plugin (DEAR, Cin7, Unleashed), $50 to $250

- A CRM (HubSpot Starter, Pipedrive), $30 to $100

- A bookkeeper for the parts none of the above handle, $400 to $800

You are at $699 to $1,369 per month before doing anything fancy. A year is $8,000 to $16,000. Every line item is rent paid to a vendor whose roadmap you do not control.

I call this the Xero stack tax. Not the $25 starting price. The six tools you accumulated to make Xero work for a business that grew past what Xero alone covers. If your monthly software bill across Xero, Payroll, Projects, an inventory plugin, and a CRM has crept past $300, this Xero alternative open source post is for you.

## Where Xero stops

Xero is excellent at accounting. The honest list of where it stops:

Inventory at any depth. Xero tracks quantity on hand and average cost. Run a Shopify store with 200 SKUs needing lot tracking, multi-warehouse transfers, or FIFO costing per item, and Xero will not get you there. The official answer is “use a Marketplace app” and you end up paying DEAR or Cin7 $99 to $250 a month.

Manufacturing. No bill of materials. No work orders. No production planning. If you assemble anything (kits, recipes, light manufacturing), Xero has no concept of it.

CRM. Not a CRM. Xero has a contacts list, but no leads, opportunities, pipelines, or sales activity tracking. People run HubSpot or Pipedrive on the side and try to keep contact data in sync.

HR beyond payroll. Xero Payroll handles paying people. No leave tracking that connects to payroll, no expense claims workflow that posts to the GL, no performance reviews. You bolt on a separate HR product or run it in a spreadsheet.

Payroll outside US, UK, AU, NZ. Mature in those four markets, weak or absent everywhere else. Canadian subsidiary or a contractor in Singapore means a separate payroll provider and a monthly CSV journal import.

Multi-entity at no extra cost. Xero charges per organization. Two LLCs means two subscriptions. Five entities means five subscriptions.

None of those are bugs. They are scope decisions. Xero decided to be the best cloud accounting product, not a full ERP. That decision is correct for what they are. It just means at some point in your growth, the rest of your business needs a different tool.

## What ERPClaw replaces beyond accounting

This is the differentiated point and the reason I am writing this Xero alternative post.

A typical “best Xero alternative” listicle compares other accounting tools: QuickBooks, FreshBooks, Wave, Zoho Books, Sage. Same shape, slightly different price. ERPClaw is a different category. It is a full ERP that includes accounting, the same way SAP includes accounting and NetSuite includes accounting. The accounting piece is one domain among many in a single system, sharing one database, with all data queryable by any module.

You stop running a stack and start running a system. ERPClaw includes:

- Accounting: US chart of accounts out of the box, audit-grade ledger, multi-company, AR/AP aging, trial balance, P&L, balance sheet, cash flow. See features.

- Inventory: items, warehouses, batches, serial numbers, FIFO and weighted-average costing. Real inventory, not “list of products with quantity on hand.”

- Manufacturing: bill of materials, production planning, work orders, routing. Does not exist in Xero at any tier.

- HR + Payroll: employees, leave, attendance, expense claims posting to GL, salary structures, FICA, federal and state income tax, W-2 generation. Same database as the GL.

- CRM: leads, opportunities, pipelines, contacts. Not Salesforce, but enough to stop paying for HubSpot Starter.

- Projects + Billing: project tracking, time entries, recurring invoices, usage-based billing.

- Stripe + Shopify deep integrations: all core Stripe and Shopify actions, three-layer payout reconciliation, ASC 606 revenue recognition. Replaces the $30 to $200 a month tools that do this. Detailed in the Stripe launch post.

- Industry verticals: retail, restaurant, healthcare, legal, nonprofit, education, real estate, and more. Install only the ones you need.

Total those as separate subscriptions plus Xero plus a Marketplace inventory plugin and you are past $1,000 per month for a 25-person company. ERPClaw is $0 forever (open source), self-hosted on your own infrastructure.

The structural piece, beyond price: when accounting, inventory, payroll, and CRM all live in one database, your AR aging report knows the customer’s CRM history, your inventory valuation knows your manufacturing routings, and your project P&L knows actual labor hours from HR. The connections are inherent, not “supported via integration.” No sync to break. No 11pm “DEAR import failed” email.

## Free and open source as a structural advantage

Xero cannot be free. Xero Limited is a public company on the ASX with thousands of employees and shareholders. Pricing has to scale with you because their cost structure scales with them. Not a moral failing, just how a public software company works.

ERPClaw can be free because none of that exists. It is built at AvanSaber Inc by a small co-founder team (me, my co-founder Varun, and a handful of advisors), pairing with Claude Code as a coding companion. Marginal cost of one more user is zero: no cloud bill (you host it), no support tier, no per-seat sales overhead. The open source license means it stays free even if the company changes hands.

The bigger consequence: scope can grow without re-pricing. When ERPClaw adds a vertical, you get it for $0. When Xero adds a feature, it is gated to a higher tier or sold as a separate plan.

In year three, the Xero bill is roughly twice what it was in year one. The ERPClaw bill is still $0.

## Side by side

The full comparison is on /compare/xero. Here is the table in this post for reference:

| Feature
| Xero
| ERPClaw

| Price
| $25 to $90/month
| Free (open source license)

| Users included
| Tier-dependent
| Unlimited (always)

| Multi-entity
| Per-organization subscription
| Unlimited entities free

| Inventory
| Add-on or premium tier
| Built-in (multi-warehouse, batches)

| Manufacturing
| No
| Built-in (BOM, MRP, work orders)

| HR / Payroll
| Add-on (Xero Payroll)
| Built-in (US payroll, FICA, W-2)

| CRM
| No
| Built-in

| AI capabilities
| AI-decorated (chat sidebar bolted on)
| AI-native (assistant is the primary interface)

| Self-hosted
| No (cloud only)
| Yes (your server)

| Open source
| Proprietary
| Open source (GPL v3)

| Database
| Xero’s cloud only
| SQLite or PostgreSQL via PyPika, on your machine

| Stripe + Shopify
| Marketplace apps required
| Built-in deep integrations

| Multi-currency
| Premium tier only
| USD, EUR, GBP, CAD, INR, SGD, AED

A few of those rows deserve a second look.

Inventory. Xero on Established has line-item inventory, average cost, and quantity on hand. Fine for a freelancer selling 10 products. For a product business with 200+ SKUs across two warehouses, you end up renting DEAR or Cin7. ERPClaw has multi-warehouse, batch and serial tracking, and FIFO and weighted-average costing in the core.

Multi-entity. Stings if you have ever opened a second Xero subscription for your holding company. ERPClaw is multi-company in the core. One install, N companies, intercompany and consolidation included.

Open source. Xero is a closed product. If they raise prices 30%, you pay or you migrate. ERPClaw is GPL v3, the code is on github.com/avansaber/erpclaw, and you can fork it and run your own version forever if you want.

## When you should stay on Xero

I promised honest. Cases where I would tell you to stay on Xero:

You are in Australia, New Zealand, or the UK and you only need accounting. Xero in those markets is a special case. Bank feed coverage, BAS/VAT/MTD integrations, and the accountant network are real advantages. If your business is “services shop with no inventory, no manufacturing, no second entity” in one of those markets, switching costs more than it saves.

Your accountant flatly refuses to work outside Xero. Some accountants work with anything that produces a clean trial balance (/docs). Some only know Xero and will charge a “switch fee” of $1,500. If yours is the second kind and you love them, the math may not work. Ask before you switch.

You will not run your own server. ERPClaw is self-hosted by default. If “I do not want to think about a server, ever” is a hard line, Xero handles that, or have us run it: ERPClaw Cloud Managed is our hosted offering, with onboarding by request through /pricing/.

You depend heavily on one specific Xero Marketplace app. If your whole inventory or payroll workflow runs through one app you have built process around, you are switching two products at once. Worth doing eventually, but not on a random Tuesday.

If none of those apply, ERPClaw is worth a weekend to evaluate.

## How a Xero migration actually works

Full step-by-step lives at /blog/migrating-from-xero-to-erpclaw (publishing Q4 2026). Short version:

- Export from Xero. Chart of accounts, contacts, inventory items, invoices, bills, general ledger. ~15 minutes of clicking.

- Map your chart of accounts. ERPClaw ships with a US default chart of accounts. Our mapping script reads your Xero export and produces a one-to-one mapping with a manual review step. 30 to 90 minutes.

- Pick a cutover date. Start of a month or quarter. Close Xero cleanly through the prior period, then open ERPClaw with opening balances as of day 1 of the new period.

- Import opening balances. One command loads your trial balance as opening entries. The validation engine rejects any imbalance.

- Import master data. Customers, suppliers, items, employees. ~15 minutes.

- Run parallel for one month. Most migration guides skip this and you should not. Post the same transactions to both for one full month, then compare trial balances. They should match to the cent.

For a typical 25-person business, one to three weekends. For a freelancer, two hours.

## FAQ

Is ERPClaw a real Xero alternative US small businesses can run, not just a side project?

It is in production use, code is on GitHub, and v1 covers full double-entry accounting plus the rest of the back-office modules. The validation engine is backed by a deep suite of constitutional, contract, and smoke tests that refuse to let regressions ship. Early but not toy software. For US small businesses, this is a genuine option today, and regional modules for the UK, EU, Canada, and India are shipped as well. For Australia and New Zealand, where those regional modules do not exist yet, Xero is still the better fit.

Will my accountant still be able to help me?

Yes if your accountant understands a trial balance and a ledger export, which is the same skill they use for non-Xero clients. ERPClaw exports a clean trial balance, P&L, balance sheet, and detailed ledger in CSV or PDF. If your accountant insists on Xero advisor write-access specifically, ERPClaw will not satisfy that. Ask before you switch.

How does Xero vs open source ERP shake out for a 10-person services business?

Xero costs roughly $90/month base plus add-ons (Projects $21, Expenses $20, possibly Payroll $80) so $210 to $270/month all in. ERPClaw is $0 forever, self-hosted. Roughly $2,500+ saved per year on software fees alone. The bigger win is consolidation: one system, one database, one place to look.

Is there a genuinely free Xero alternative for one-person shops?

Yes. ERPClaw is free at any size. Wave is also free if you only need US accounting and invoicing. ERPClaw covers the same ground plus inventory, projects, payroll, CRM, and the rest. For anyone who might add inventory, employees, or a second entity, ERPClaw scales further at the same $0.

Does ERPClaw handle multi-currency the way Xero Premium does?

ERPClaw supports USD, EUR, GBP, CAD, INR, SGD, and AED. A payment settles in the currency its invoice was raised in. What ERPClaw does not do is convert between currencies inside a transaction, which is what Xero Premium’s automatic gain and loss tracking is for; if that specific behavior is a hard requirement, Xero keeps the edge there.

What about payroll outside the US?

ERPClaw v1 payroll is US-focused: FICA, federal and state income tax for all 50 states, W-2 generation. UK PAYE, AU STP, NZ PAYE are not there yet. Xero Payroll wins in those markets today.

Of the Xero competitors 2026 buyers are looking at, why ERPClaw vs ERPNext or Odoo?

ERPNext and Odoo are also open-source ERPs with a similar pitch on paper. The differences: ERPClaw is AI-native (chat is the default interface, not bolted on top of forms), ships with deep Stripe and Shopify integrations in the core, and is built for US small businesses first rather than mid-market global. Compare at /compare/quickbooks and /compare/erpnext.

## Try it

Install takes five minutes. The whole ERP, on your machine.

`clawhub install erpclaw`
That puts the core ERP (full accounting, inventory, HR, payroll) on your server. From there, talk to it: “I run a 12-person retail shop in Portland, set me up.” It picks the industry vertical, installs the relevant modules, generates the chart of accounts, and is ready for transactions.

If you want to read first:

- /features for the full module breakdown

- /pricing (spoiler: it is free)

- /compare/xero for the side by side

- /compare/quickbooks for the QuickBooks version of this post

- /for/small-business for the SMB pitch

- /docs for install and migration walkthroughs

Repo: github.com/avansaber/erpclaw. Star it to follow along, we ship weekly. The Xero App Marketplace is worth a look if you want to compare an “extend Xero” approach next to the “one-system” approach.

Xero is the right software for what it is: the best cloud accounting product for a small business that only needs accounting, especially in Australia, New Zealand, and the UK. ERPClaw is the right software when accounting is one of the things you need and you are tired of renting the other six. If you are running an inventory plugin, paying for Xero Projects, paying for Xero Payroll, and bolting on a CRM, you crossed that line a while ago.

Time to install.

---

## Acumatica Alternative: Free AI-Native ERP
URL: https://www.erpclaw.ai/compare/acumatica/

# ERPClaw vs Acumatica
Acumatica is mid-market cloud ERP with resource-based pricing. ERPClaw is the open-source AI-native alternative at $0 forever.
Same primitives. AI-native by design. Free.

| Feature | Acumatica | ERPClaw
| Pricing model | Resource-based, contact sales | $0 forever
| Implementation cost | $50K-300K typical | $0 (5-min install or self-migrate)
| Implementation time | 3-9 months | 5 minutes
| Modules included | By edition; cross-sell expected | All 46 included always
| Multi-entity | Yes (mid-market editions) | Yes
| AI capabilities | Acumatica AI add-ons (decoration on a 2008 ERP) | AI-native architecture
| Customization | Acumatica xRP framework, .NET/C# | Spec-first; any LLM regenerates code
| Self-hosted | Yes (private cloud option) | Yes
| Open source | Proprietary | Free &amp; Open Source
| Vendor lock-in | Significant (proprietary stack) | None (cp data.sqlite to anywhere)

## Key differences

### $0 vs resource-based pricing
Acumatica's resource-based pricing avoids per-seat creep but the all-in cost still lands $50K-$300K per year for mid-market deployments. ERPClaw is $0 forever, every module, with a 5-minute install.

### AI-native instead of AI add-on
Acumatica added AI features to a 2008 ERP. ERPClaw was rebuilt around AI. The 5-trait test on /ai-native-erp/ separates the architectures.

### Spec-first regen vs xRP framework
Acumatica customizations live in xRP and .NET. ERPClaw customizations live in SKILL.md specs that any LLM regenerates from. No partner engagement needed.

### Open source vs proprietary
Acumatica is closed-source even when self-hosted; you cannot read or modify the underlying code. ERPClaw is Free &amp; Open Source with full source on GitHub.

## Frequently asked questions

### How does Acumatica pricing compare to ERPClaw?
Acumatica uses resource-based pricing (transactions, GB of data, users in some configurations), and the published number is opaque because it lands at contact sales. Mid-market deployments commonly land $50K to $300K per year all-in. ERPClaw is $0 forever, with all modules included and no per-resource metering.

### Is Acumatica really self-hosted like ERPClaw?
Acumatica offers a private-cloud deployment option, but the software remains proprietary; you self-host the closed-source binary. ERPClaw is Free &amp; Open Source, so self-hosting includes the right to read, modify, fork, and audit the source.

### What about Acumatica's xRP customization framework?
xRP is Acumatica's customization platform built on .NET/C#. Customizations are real engineering engagements with implementation partners. ERPClaw's spec-first regen pattern lets any LLM generate customization code from a SKILL.md spec, which collapses the partner cost.

### Is Acumatica AI-native?
No. Acumatica added AI features to its existing forms-and-workflows ERP. The architecture predates the AI agent era. ERPClaw was AI-native by design from line one. The 5-trait test on /ai-native-erp/ scores both layer by layer.

### When would I still pick Acumatica over ERPClaw?
If you are at the upper mid-market edge with deeply specialized industry SuiteApps that ERPClaw's 14 verticals do not cover, Acumatica's larger ISV ecosystem can be worth the cost. For most SMB-to-mid-market deployments, ERPClaw covers the same primitives at $0 with structurally newer architecture.

---

## ERPClaw vs the AI-ERP Stack You'd Otherwise Assemble
URL: https://www.erpclaw.ai/compare/ai-erp-stack/

# ERPClaw vs the AI-ERP stack you'd otherwise assemble
The 2026 AI finance market sells excellent slices: a close product here, an accounting service there, an analytics layer, procurement agents. Each slice still needs the others. ERPClaw ships the whole spine in one free, self-hosted, AI-native install.
By Varun Borawake, Co-founder, ERPClaw · Published 2026-08-14 · Vendor facts checked 2026-08-14

## The stack, slice by slice
Assemble the current AI finance market into one company's tooling and this is what it looks like. Every vendor description below comes from that vendor's own public materials, checked 2026-08-14.
| The job | Who sells it as a slice | The catch | In ERPClaw
| General ledger and close | Campfire, Rillet, DualEntry, or Puzzle | Closed SaaS; pricing demo-gated except Puzzle ($25 to $300 per month, published) | Accounting, close, multi-entity consolidation, reports by any dimension
| AI accountants running the close | Billow | Service on top of your existing ERP; pricing not public | Not needed: the invariant engine keeps the books closeable continuously
| Analytics and plain-English answers | Definite | Copies your data into its platform; free tier, then $250 per month and up | Questions run against the ledger itself, no copy, no meter
| Procurement agents | Spaceflow | Agents across your existing ERP; pricing not public | Purchasing, RFQs, and procure-to-pay in the same books
| Inventory and manufacturing | Integrations (e.g. Campfire routes to Cin7; DualEntry offers none) | Separate subscriptions or a gap | Stock with hard reservations, putaway, pick lists, work orders, subcontracting
| Payroll | Integrations (ADP, Gusto, Rippling) across the cohort | Separate subscription | US payroll in the same install
| Sales pipeline | A CRM subscription | Separate subscription | Built-in Sales from first lead to closed order

Deep dives per vendor: Campfire · Rillet · DualEntry · Puzzle · Billow · the pattern itself: AI-native vs the AI layer

## Why the seams are the product
Every pair of slices in that stack meets at an integration: the close product syncs from the billing system, the analytics layer copies from both, the procurement agents write back under supervision, and payroll lives in a fourth vendor's cloud. Each seam is a place where two systems can disagree about the truth, and reconciling those disagreements is where finance teams actually spend their hours.
ERPClaw's answer is structural, and it is the same AI-native architecture the best of these companies bet on, carried all the way down: one shared database, every operation invokable in plain language, deterministic code doing the posting, and an invariant engine checking the books after every entry. When the deal, the delivery, the payslip, and the ledger entry are one system, there are no seams to reconcile and no slice left to buy.
And the economics are not close. The stack's bills stack; several of them will not even tell you the number without a call. ERPClaw is open source and free forever, self-hosted, PostgreSQL or SQLite. The suite argument in one sentence: each of them sells a slice that still needs the others; we ship the whole spine, one install, $0.

## Common questions

### Is this comparison fair? These companies do not claim to be full ERPs.
Mostly they do not, and that is the point. Each sells one slice of what a business runs on, and each slice is genuinely good at its job. The comparison is not slice versus slice; it is the assembled stack, with its bills, syncs, and seams, versus one free install that ships the spine end to end.

### What does the assembled stack cost?
Mostly unknowable in advance, which is itself the finding: of the vendors named here, only Puzzle and Definite publish prices. The rest are demo-gated or unpublished, sitting on top of whatever the incumbent ERP already costs. ERPClaw's number is $0 forever, published, for the whole spine.

### Where is ERPClaw honestly weaker than a best-of-breed slice?
Polish per slice. A purpose-built close cockpit like Rillet's or Campfire's is a more refined screen for that one ritual; ERPClaw closes the books inside the ERP that runs everything else. Vendor-run migration services are real conveniences too. What the slices cannot offer at any price is one database where the deal, the stock, the payslip, and the ledger entry are the same system, on your hardware.

### Can I mix ERPClaw with these tools?
Yes. ERPClaw is open source with the action layer as its API, and nothing about self-hosting prevents a layer tool reading from it. The dependency does not run the other way: ERPClaw does not need any of them to be complete.

### Why is 'one shared database' worth so much?
Because every seam in a stack is a place where two systems disagree about the truth, and reconciling those disagreements is where finance teams' hours actually go. When there is one set of books and everything posts into it under invariant checks, the reconciliation industry around your own data disappears.

## One spine, no seams, no bill
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawBook a walkthrough

---

## Billow Alternative: Own the Ledger Instead
URL: https://www.erpclaw.ai/compare/billow/

# Billow alternative: own the ledger instead of renting help for it
Billow (thebillow.ai) puts AI accountants on top of the ERP you already pay for. ERPClaw is the AI-native ERP itself: open source, self-hosted, $0 forever. One manages the pain of a legacy ledger; the other replaces the ledger.
By Varun Borawake, Co-founder, ERPClaw · Published 2026-08-14

Billow AI Labs, a Y Combinator summer 2026 company, describes itself as an "AI-native Accounting Firm" whose system "runs your close, reconciliations, and reporting" while "you review the exceptions." It connects to NetSuite, QuickBooks, Xero and a long list of finance tools, reads your books, and posts journal entries into them. Pricing is not public.
Notice what the model presumes: an incumbent ERP stays in place, its subscription keeps running, and a second bill arrives for making it bearable. That is the "AI layer" pattern, and it is a rational response to legacy systems. The question a buyer should ask first is different: why does the ledger need a staffed layer at all?
ERPClaw's answer is architectural. When the system of record is AI-native from the start, plain-language requests as the interface, deterministic accounting code doing every posting, an invariant engine checking the books after each one, the monthly close stops being a crisis that justifies outsourcing. The layer and its bill become unnecessary together.
The full argument, with the whole layer-versus-native cohort mapped, is in AI-native vs the AI layer and the suite roll-up comparison.

## TL;DR comparison
| Dimension | Billow | ERPClaw
| Type | AI accounting service layered on your existing ERP | Open-source AI-native ERP, the system of record itself
| What you keep paying for | Your NetSuite or QuickBooks subscription stays, plus Billow on top | Nothing. The ERP is $0 forever, self-hosted
| Who writes in your books | Reads your books and posts journal entries into them | Is your books; deterministic code posts, an invariant engine checks every entry
| Pricing | Not public | $0 forever, open source
| Delivery | Cloud service; their team plus AI runs your close | Self-host (Docker, CLI, SQLite or PostgreSQL)
| Best for | Teams that want an accounting firm relationship rather than software | Businesses that want to own the ledger and the whole operation

## When Billow is the right call

### You want an accounting firm's shape, not a software purchase.
Billow positions as an AI-native replacement for outsourced accounting: their system runs reconciliations and posts entries, humans review exceptions, and compliance framing (SOC 2 Type II, SOX) is front and center. If you are buying a service relationship, that is a different aisle from software, and it is the aisle they are in. And if you want that kind of relationship with ERPClaw underneath instead, our accounting partners can help; ask through the support page.

## When ERPClaw is the right call

### You are keeping NetSuite, and the close is the pain.
That case is exactly what ERPClaw Insights covers, at $0. Keep NetSuite; Insights reads it, answers in plain language, and the per-business auto-sync posts entries back, so the monthly pain shrinks without hiring anyone's AI accountants. The pain of a legacy close does not need a staffed service; it needs better software beside the ledger.

### You would rather delete the problem than outsource it.
Billow exists because old systems make the monthly close painful enough to hire help for. ERPClaw removes the pain at the source: you ask in plain language, real accounting code does the posting, and the books are checked after every entry. Books kept that way do not turn into a monthly crisis, so there is nothing to outsource.

### You are not on NetSuite yet, and would rather never be.
Billow's model presumes an incumbent ERP to layer onto, and their integration list starts at NetSuite, QuickBooks, and Xero. If you have not signed that subscription yet, the cheaper move is an ERP that needs no layer: $0 forever, self-hosted, whole-business scope.

### You want to know what things cost before a sales call.
Billow publishes no pricing. Under it, you are still paying for the ERP it manages. ERPClaw's total software bill is public: $0, with the books, inventory, purchasing, payroll, and Sales included.

### You don't want anyone else writing in your books, ever.
Billow's service posts journal entries into your books, with your team reviewing exceptions. That is a reasonable design for what it is, and it is still a third party with write access to your ledger. In ERPClaw the books live on your machine, every posting runs through the same invariant checks, and nobody has write access you did not grant on your own hardware.

## What it costs
| Item | Billow | ERPClaw
| The service | Not public (checked 2026-08-14) | No service needed; the product is the answer
| The ERP underneath | Your existing NetSuite, QuickBooks, or Xero bill continues | $0 forever, unlimited users
| Source code | Closed; it is a service | Free &amp; Open Source: fork it, audit it
| Where your books live | In the incumbent ERP's cloud, written to by the service | On your hardware, written only by you

## Common questions

### Is Billow a competitor to ERPClaw at all?
Structurally, they are answers to the same pain from opposite directions. Billow (thebillow.ai, Billow AI Labs) layers AI accountants on top of the ERP you already pay for. ERPClaw replaces that ERP with an AI-native one you own for $0. If the close hurts, one option manages the hurt and the other removes its cause.

### What is Billow exactly?
A Y Combinator summer 2026 startup founded in 2026, positioning as an AI-native accounting firm: their system connects to platforms like NetSuite, QuickBooks, and Xero, runs reconciliations and close work, posts entries, and routes exceptions to humans. Pricing is not public.

### Does ERPClaw have something like Billow's AI accountants?
The equivalent capability lives in the product rather than a service: plain-language operation of the whole ledger, automated matching on bank statement import, and an invariant engine that refuses postings that would break the books. You operate it yourself, which is the point.

### Could someone run Billow on top of ERPClaw?
Nothing stops an accounting firm from operating ERPClaw for clients; it is open source and free, and firms do exactly that with self-hosted software. But the AI-native design exists to make the books operable by the business itself, in its own words.

### Who should honestly pick Billow?
A team that specifically wants an accounting firm relationship: humans accountable for the close, bought as a service. Keeping NetSuite is not by itself a reason, since ERPClaw Insights covers keep-your-system at $0, and a service relationship around ERPClaw is available through our accounting partners. If ownership of the ledger matters at all, that is ERPClaw's side of the table.

## Fix the books once, and you won't need the extra service
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawKeep your books, get the answers: Insights

---

## Campfire Alternative That's Free and Self-Hosted
URL: https://www.erpclaw.ai/compare/campfire/

# Campfire alternative: ERPClaw is open-source, whole-business, and free
Campfire (campfire.ai) and ERPClaw are both AI-native. Campfire is closed SaaS focused on the finance suite, with payroll, inventory, and procurement handled by integrations. ERPClaw is the open-source AI-native ERP: the books plus the rest of the business, self-hosted, $0 forever.
By Varun Borawake, Co-founder, ERPClaw · Published 2026-08-14

Campfire calls itself "the AI-native ERP that closes your books faster," and unlike most products wearing that label, the claim holds up: founded 2023, Y Combinator summer 2023, then a $35M Series A led by Accel in June 2025 and a $65M Series B co-led by Accel and Ribbit in October 2025, all on an AI-native architecture with their Ember assistant at the center. This is a serious product with real customers named on their site.
So the comparison is not "real AI versus fake AI." It is three structural questions, and they have different answers than they do for legacy vendors:

- Scope. Campfire's modules are the finance suite; their own integrations page routes payroll (ADP, Gusto, Rippling), inventory (Cin7, DOSS), and procurement (Ramp, Brex, BILL) to other vendors. ERPClaw builds those into the same database as the ledger.
- Delivery. Campfire is their cloud, full stop. ERPClaw runs on your machine, SQLite or PostgreSQL, source open under Free & Open Source.
- Economics. Campfire's pricing is demo-gated, with no public figures anywhere (checked 2026-08-14). ERPClaw costs $0 forever, and that is the architecture, not a promotion.
For the wider map of where Campfire, Rillet, Puzzle, and ERPClaw all sit, start with our AI-native ERP framework and the suite roll-up comparison.

## TL;DR comparison
| Dimension | Campfire | ERPClaw
| Type | Commercial AI-native ERP (finance suite) | Open-source AI-native ERP (whole business)
| Delivery | Vendor SaaS only | Self-host (Docker, CLI, SQLite or PostgreSQL)
| AI focus | Ember assistant: close, reconciliations, categorization, anomaly flags | Natural-language action layer across every business operation
| Pricing | Not public, demo-gated | $0 forever, open source
| Scope edges | Payroll, inventory, and procurement are integrations, not modules | Inventory, manufacturing, purchasing, payroll, and Sales built in
| Best for | Mid-market SaaS finance teams leaving NetSuite for managed close | Owners and technical teams who want the whole business in one system they control

## When Campfire is the right call

### You want a vendor-run SaaS and no ops at all.
Campfire is cloud software run by Campfire: no server, no updates to apply. That preference alone no longer decides the table, because ERPClaw Cloud Managed is our vendor-run option, onboarding by request through /pricing/. Where Campfire still wins the point is the dedicated vendor support team a larger buyer may want on contract.

### Finance close for a software company is the whole job.
Campfire's product is built around the close: checklists, flux analysis, continuous reconciliation, board-ready statements, with customers like Replit and PostHog named on their site. If your business is SaaS and your pain is days-to-close, their product fit is direct.

### You are mid-market and want the vendor migration service.
Campfire's team performs transaction-level historical migrations out of NetSuite for you, and their case studies lean on that. If you want to hand the move to a vendor rather than run it, that service is the product.

### Your stack is already the one they integrate.
Payroll through ADP, Gusto, or Rippling, spend through Ramp or Brex, inventory through Cin7: if those subscriptions are staying anyway, Campfire's integration-first shape may fit how you already operate.

## When ERPClaw is the right call

### You want the functions they integrate to be built in.
Campfire's own integrations page routes payroll, inventory, and procurement to other vendors' products. In ERPClaw, inventory, manufacturing, purchasing, payroll, and built-in Sales live in the same database as the books. One system, one truth, no per-module subscriptions to other companies.

### You want to see the price before the demo.
Campfire publishes no pricing; their pricing URL does not even resolve. ERPClaw's price is public and permanent: $0 forever, unlimited use, open source. The comparison is one sentence long.

### You want your books on your own machine.
Campfire is SaaS only; no self-hosted option is documented anywhere on their site. ERPClaw is self-hosted by design, SQLite or PostgreSQL, and your ledger is a database file you can open, back up, and audit yourself.

### You want AI-native across the whole business, not the finance slice.
Both products are AI-native, and that matters: neither is a chat window bolted onto menus. The difference is reach. Ember answers accounting questions and drafts entries. ERPClaw's plain-language layer runs the whole company: invoices, stock reservations, work orders, payroll, and the close, with deterministic code posting and an invariant engine checking every entry.

### You want out to always be open.
Campfire's documented data-out paths are its API and CSV endpoints. ERPClaw does not have an export story because it does not need one: the data never leaves your possession in the first place, and the source code is Free &amp; Open Source.

## What it costs
| Item | Campfire | ERPClaw
| Subscription | Not public; demo-gated (checked 2026-08-14) | $0 forever, unlimited users
| Implementation | Vendor-run migration service | Install in minutes, file-based imports documented
| Source code | Closed | Free &amp; Open Source: fork it, audit it
| Where it runs | Their cloud only | Your hardware, your choice of database

## Common questions

### Is Campfire AI-native or AI-decorated?
AI-native, by their own architecture and ours: Campfire was built in the AI era around their Ember assistant and automated reconciliation, and raised over $100M on that thesis. This comparison is between two AI-native products; the wedge is scope, delivery, and economics, not whether the AI is real.

### What does Campfire cost?
They do not say publicly. There is no pricing page (the URL returns a not-found error, checked 2026-08-14), so budget conversations start with a sales call. ERPClaw is $0 forever, which you did not need a call to learn.

### Can ERPClaw replace Campfire for a SaaS company's close?
The accounting engine, multi-entity consolidation, revenue side, and bank reconciliation are all there, operated in plain language. Campfire's purpose-built close cockpit is the more refined screen for that one ritual today; ERPClaw closes the books inside the ERP that runs everything else, with automated close set up per business. If days-to-close is your only metric, evaluate both side by side. If the rest of the business needs running too, the scope difference decides it.

### What about inventory and payroll?
That is the clearest structural difference. Campfire hands inventory to integrations like Cin7 and payroll to ADP, Gusto, or Rippling, which means more vendors and more bills. ERPClaw ships inventory with hard reservations, manufacturing with subcontracting, and US payroll in the same install as the ledger.

### Who is Campfire actually for?
By their own site: mid-market and enterprise software companies, often moving off NetSuite, who want a managed vendor relationship. If that is you and self-hosting is a non-starter, they are a credible option. ERPClaw is for the buyer who wants ownership: of the data, the code, and the whole operation.

## See the whole-business version
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawBook a walkthrough

---

## DualEntry Alternative With Inventory Built In
URL: https://www.erpclaw.ai/compare/dualentry/

# DualEntry alternative: the AI-native ERP that also runs your inventory
DualEntry (dualentry.com) and ERPClaw are both AI-native. DualEntry is a closed, cloud-only financials suite with no inventory, manufacturing, HR, or payroll. ERPClaw is the open-source AI-native ERP that runs the whole business, self-hosted, $0 forever.
By Varun Borawake, Co-founder, ERPClaw · Published 2026-08-14

DualEntry markets itself as "the AI ERP that just works": founded 2024 in New York, a $6M seed in May 2024 led by Contrary, then a $90M Series A led by Lightspeed and Khosla Ventures announced in October 2025. The AI-native claim is genuine, and their NetSuite-switcher pitch, data migrated in 24 hours with zero implementation fees, is the sharpest piece of marketing in the category.
The comparison turns on what the word "ERP" is doing in each product's name:

- Scope. DualEntry's ERP is financials: GL, AR, AP, revenue recognition, consolidation, planning, treasury. No inventory, no manufacturing, no warehouse, no HR, no payroll. ERPClaw's ERP includes all five, in the same database as the ledger.
- Delivery. DualEntry is their cloud only. ERPClaw runs on your hardware, SQLite or PostgreSQL, Free & Open Source.
- Economics. DualEntry names tiers and withholds every price, with no free trial. ERPClaw is $0 forever, publicly, permanently.
For the wider category map, start with our AI-native ERP framework and the suite roll-up comparison.

## TL;DR comparison
| Dimension | DualEntry | ERPClaw
| Type | Commercial AI-native ERP (financials only) | Open-source AI-native ERP (whole business)
| Delivery | Vendor SaaS only, North American data centers | Self-host anywhere (Docker, CLI, SQLite or PostgreSQL)
| AI focus | Categorization, reconciliation matching, anomaly flags, copilot that drafts entries | Natural-language action layer across every business operation
| Pricing | Demo-gated, three named tiers, no dollar figures, no free trial | $0 forever, open source
| Scope edges | No inventory, manufacturing, warehouse, HR, or payroll | All of those built in, same database as the books
| Best for | Mid-market finance teams replacing NetSuite with managed SaaS | Businesses that make, stock, or ship things, and anyone who wants ownership

## When DualEntry is the right call

### You are a financials-only mid-market company that wants managed SaaS.
DualEntry's module list is deep on finance: multi-entity consolidation, multi-book, revenue recognition, planning, treasury. If your business is services or software and your team lives in the GL, their product is aimed squarely at you. One correction to the usual framing, though: self-hosting being off the table is not a reason on its own, because ERPClaw Cloud Managed is our vendor-run option, onboarding by request through /pricing/.

### The vendor-run migration is the deciding factor.
Their NextDay Migration pitch is 'migrate in 24 hours, zero implementation fees,' with transaction-level history from NetSuite, Sage Intacct, QuickBooks and others. Read their own fine print, the data moves in 24 hours and go-live is four to six weeks, but a vendor doing the move for you at no charge is a real offer.

### You want a funded vendor with mid-market references.
A $90M Series A led by Lightspeed and Khosla in October 2025, named customers on their site, and unlimited users on every tier. If your board wants a venture-backed vendor with a support contract, that is what this is.

## When ERPClaw is the right call

### Your business has things, not just entries.
This is the whole comparison for many buyers. An independent ERP reviewer put it plainly in July 2026: DualEntry has no inventory, manufacturing, warehouse, or field-service capability of any kind, and their own module list confirms the shape. If you stock, make, or ship anything, ERPClaw runs inventory with hard reservations, manufacturing with work orders and subcontracting, and purchasing, in the same books.

### You want people functions in the same system.
DualEntry offers no HR or payroll. ERPClaw ships both, alongside built-in Sales from first lead to closed order, so the deal, the delivery, the payslip, and the ledger entry live in one database.

### You want the price without the demo.
DualEntry names three tiers and prices none of them, and there is no free trial. ERPClaw's number is public and permanent: $0 forever. The evaluation costs you an afternoon, not a sales cycle.

### You want your books in your own hands.
DualEntry is cloud-only in their data centers. ERPClaw is self-hosted by design: your machine, SQLite or PostgreSQL, source code open under Free &amp; Open Source, and an invariant engine checking every posting. Both are AI-native; only one of them is yours.

### You like the CLI idea, but want it to run the whole company.
DualEntry markets 'the first ERP with a CLI,' which tells you who they think the modern buyer is. ERPClaw agrees and goes further: the plain-language layer, CLI included, is the primary interface for every operation in the system, not a companion to a finance web app.

## What it costs
| Item | DualEntry | ERPClaw
| Subscription | Demo-gated; tiers named, figures withheld; extra users billed per user per month | $0 forever, unlimited users
| Implementation | Included; go-live in 4 to 6 weeks per their own page | Install in minutes, file-based imports documented
| Free trial | None, per their FAQ | The product is free; the trial is the product
| Source code | Closed | Free &amp; Open Source: fork it, audit it

## Common questions

### Is DualEntry AI-native or AI-decorated?
AI-native. They were founded in 2024, describe themselves as built from the ground up with AI, and their features, learned categorization, matching, anomaly detection, a copilot that drafts entries, are architectural, not bolted on. This page compares two AI-native products; the wedge is scope, delivery, and economics.

### What does DualEntry actually cost?
They do not say publicly. Three tiers are named with no dollar amounts, additional users cost an unstated per-user monthly fee, and there is no free trial (their own FAQ, checked 2026-08-14). ERPClaw is $0 forever.

### Can DualEntry run inventory or manufacturing?
Their module list contains neither, and a third-party ERP reviewer stated in July 2026 that it has no inventory, manufacturing, warehouse, or field-service capability of any kind. ERPClaw ships all of it: stock with hard reservations, putaway and pick lists, work orders, subcontracting, and landed costs.

### Their migration says 24 hours. Is that real?
Their own pages say both things: data migration in 24 hours, go-live in four to six weeks. Quote them precisely and plan for the second number. ERPClaw's file-based migration is documented, self-serve, and free, because there is no implementation invoice in a $0 product.

### Who should honestly pick DualEntry?
A financials-only mid-market company that wants managed SaaS, vendor support, and a done-for-you NetSuite exit, and does not mind closed source or unpublished pricing. If any part of your business touches physical goods, or you want ownership of code and data, that is ERPClaw's side of the table.

## The ERP where "everything" includes the warehouse
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawERPClaw for Manufacturing

---

## ERPNext Alternative - AI-Native Open Source ERP
URL: https://www.erpclaw.ai/compare/erpnext/

# ERPClaw vs ERPNext
Both are open source ERPs. ERPNext is a traditional web application. ERPClaw is AI-native and runs from your chat.

| Aspect | ERPClaw | ERPNext
| Price | Free (open source license) | Free (GPL) or $50+/mo cloud
| Interface | Natural language chat | Traditional web UI (Desk)
| Setup time | 5 minutes | Hours to days
| Training | None needed | Significant learning curve
| Infrastructure | SQLite (default, single file) or PostgreSQL | MariaDB + Redis + Node.js + nginx
| Server requirements | 2 GB RAM, any Linux/macOS | 4+ GB RAM, specific stack
| Database backup | Copy a file | mysqldump + redis backup
| AI integration | AI-native (built into the action layer) | Third-party apps (NextAI, ChatNext, ChangAI). Plus an MCP option via Composio.
| Framework dependency | None (pure Python + SQLite or PostgreSQL via PyPika) | Frappe framework required
| Deployment | Copy files to server | bench setup + bench install
| Customization | Edit Python scripts directly | DocType system + hooks
| Multi-tenant | Multi-company in single DB | Separate sites per tenant
| License | Free &amp; Open Source (open source) | Free &amp; Open Source (open source)
| Offline capability | Always (local SQLite) | Limited

ERPNext AI ecosystem facts verified 2026-05 from publicly available Frappe marketplace and GitHub documentation.

## Different Philosophies

### Chat vs. Forms
ERPNext provides a traditional web UI with DocTypes and forms. ERPClaw eliminates the UI entirely and lets you interact through natural language on any messaging platform.

### SQLite (or PostgreSQL) vs. MariaDB Stack
ERPNext requires MariaDB, Redis, Node.js, and nginx. ERPClaw is database-agnostic via PyPika: SQLite by default (single file, no server to install), PostgreSQL fully supported for teams that need it. The stack complexity breakdown walks both installs side by side.

### Open source, copyleft
Both ERPClaw and ERPNext are Free & Open Source. The wedge isn't license; it's architecture. ERPNext is forms-and-workflows with AI plug-ins; ERPClaw is AI-native by design with the action layer as the API.

### Modular Skills vs. Monolith
ERPNext is a monolithic application on the Frappe framework. ERPClaw is a suite of 48 independent modules that you can install selectively. Need just accounting? Install just accounting.

### Static vs. Self-Improving
ERPNext has static modules that change only when developers push updates. ERPClaw OS detects gaps, generates new modules from 7 proven patterns, and validates every change against constitutional articles. Your ERP evolves with your business.

## Where ERPNext excels
To be fair, ERPNext has advantages in specific areas:
Mature web UI with reports, dashboards, and print formats built in

Large community and ecosystem of third-party apps

Wider international localization (ERPClaw covers US, India, Canada, UK, EU)

More established track record with large deployments

Built-in website builder and e-commerce via Frappe

## Choose ERPClaw if you want...
An AI-first experience where you manage your business through conversation

Minimal infrastructure: no MariaDB, no Redis, no nginx to manage

5-minute setup instead of hours of bench configuration

open source license freedom for commercial use without copyleft restrictions

A US-focused ERP with proper US GAAP, sales tax, and 1099 compliance

## Try it yourself
Free, open source, and ready to use. Set up in 5 minutes.
Get Started Free See All Features

Related: read the AI-native ERP framework, the open-source AI accounting map (where ERPNext sits as the AI-decorated default), the Odoo comparison, or the long-form ERPNext post. Ready to move? The ERPNext migration guide covers export, import, and cutover.

## Frequently asked questions

### Why pick ERPClaw over ERPNext when both are Free &amp; Open Source?
ERPNext is mature and has a 15-year track record but its architecture predates the AI era. The core ships without native AI; teams add it through third-party Frappe apps like NextAI, ChatNext, or ChangAI. ERPClaw was built AI-native from line one: action layer, constitutional invariants enforced at write time, spec-first regen. For workflows where the AI agent is the primary user, ERPClaw is structurally easier.

### Does ERPNext have built-in AI in 2026?
Not in the core. ERPNext's AI surface is a marketplace of third-party Frappe apps including NextAI, ChatNext, ChangAI, and Aerele's chatbot, plus an MCP integration via Composio. Each is a separate install with its own license, support, and documentation. ERPClaw bakes the AI into the action layer from line one, so every action is callable by an AI agent without a separate install.

### Is ERPNext install really a multi-hour setup?
Self-hosted ERPNext runs on the Frappe Framework stack: Python, Vue, MariaDB, Redis, Node.js, and Nginx, each with version pinning. Multiple reviewer reports describe multi-hour installs and recurring server crashes during scaling. ERPClaw is a single SQLite file with WAL mode and a 5-minute install; PostgreSQL works the same way for teams that need it.

### How is ERPClaw documentation different from ERPNext?
ERPClaw uses SKILL.md specs per module, designed for AI agents and humans together. Each module documents every action, schema, and invariant in 300 lines or fewer. ERPNext documentation has been criticized as inconsistent and forum-driven. The SKILL.md approach also means an AI assistant can answer ERPClaw documentation questions directly from the spec.

### Can I migrate from ERPNext to ERPClaw?
ERPNext exports cleanly to CSV per DocType. ERPClaw imports through the same action layer the AI agent uses. The migration script is straightforward for the foundation modules; vertical-specific schema differences need module-level mapping. We're happy to help on a real migration; open a GitHub issue.

### Does ERPClaw work for big enterprises like ERPNext doesn't?
Honest answer: not yet at the Fortune 500 scale. ERPClaw targets solo founders to mid-market, not multi-thousand-user deployments. ERPNext has the same ceiling, by reviewer reports. Both are best for small to mid-sized businesses. ERPClaw's AI-native architecture and SQLite/PostgreSQL flexibility make it a stronger pick at that tier.

---

## ERPNext Stack vs ERPClaw: 5-min Install vs Frappe Bench
URL: https://www.erpclaw.ai/compare/erpnext/stack-complexity/

# ERPNext Stack Complexity vs ERPClaw: The Install Story
ERPNext is Free & Open Source and capable. The install story is its weakest point.

## The ERPNext stack in 2026
Seven moving parts, all orchestrated by Bench. Reviewer reports put a typical install at 2-6 hours, and recurring server crashes during scaling are a known pain point.

### Frappe Framework
Application framework (Python + JS). Defines DocTypes, hooks, permissions.

### Python 3.10+
Runtime for the Frappe app and ERPNext modules.

### MariaDB or PostgreSQL
Primary relational store. MariaDB is the default.

### Redis
Caching, queues, real-time pub-sub. Multiple instances in production (cache, queue, socketio).

### Node.js
Frontend build pipeline plus the socketio server for real-time updates.

### Nginx
Reverse proxy, static asset serving, multi-site routing.

### Bench
CLI orchestrator that wires all of the above together.

## The ERPClaw stack in 2026
One Python application. Database-agnostic via PyPika. SQLite default, PostgreSQL fully supported. 5-minute install.

### Python 3.10+
Runtime for the ERPClaw action layer and every module.

### SQLite (default) or PostgreSQL
Database-agnostic via PyPika. Same code runs on either.

## Why this matters day to day
The operational cost of 5-7 services vs 1 shows up in backup, upgrade, and failure modes, not in the demo.
| Dimension | ERPNext | ERPClaw
| Number of services to run | 5-7 (MariaDB, multiple Redis, Node socketio, nginx, bench worker, scheduler) | 1 (a Python process)
| Backup | DB dump plus site folder plus private files; per-site if multi-tenant | Copy one file (SQLite) or pg_dump (PostgreSQL)
| Upgrade | bench update, migrate DocTypes, restart workers, watch for hook regressions | Replace the binary, run migrations once
| Failure modes | Any of 5-7 services can fail independently. Redis OOM and worker hangs are common. | One process; failure modes are bounded to it.
| Install time | 2-6 hours per reviewer reports; longer on first try | 5 minutes

## What ERPNext does better
Honest acknowledgment. ERPNext has been at this for longer and has earned some real wins.

### Mature DocType system
Frappe's DocType model is one of the most mature low-code metadata systems in the open source ERP space. Custom fields, custom doctypes, scripts, and workflows are first-class.

### Multi-tenant by design
Bench's multi-site model lets you run many isolated ERPNext sites on one host. The per-site database split is real isolation, not row-level multi-tenancy.

### Large existing community
ERPNext has an active community, regular conferences, and a long catalog of partner apps. The Frappe app store covers many vertical needs.

## What ERPClaw does better
The three structural moats relative to ERPNext.

### Single-binary deployment
One Python application, one database file (or one Postgres database). No service orchestration. No bench. No site folders.

### AI-native architecture
AI is built into the action layer, not bolted on through third-party Frappe apps. ERPNext's 2026 AI surface is third-party apps like NextAI, ChatNext, and ChangAI. ERPClaw's is the same action layer the human and the API call.

### GL invariants enforced pre-write
Every GL posting passes the full invariant validation before the transaction commits. Immutable GL by design. The AI agent cannot post an invalid entry.

## Frequently asked questions

### Why does ERPNext need so many services?
Frappe was built around a multi-tenant SaaS-style architecture: MariaDB or PostgreSQL for the relational store, multiple Redis instances for cache and queue and socketio, Node for the real-time layer, Nginx for routing across sites, and bench to orchestrate it all. The split is a tradeoff: it scales horizontally and isolates tenants well, at the cost of operational surface. For a single-tenant SMB install, that surface is overhead.

### Can ERPNext run on a smaller stack?
Officially, no. The supported install requires the full Frappe Bench setup. There are community attempts at minimal Docker-Compose deployments, but they still run the same set of services in containers. The component count does not shrink, only the way you start them.

### Does ERPClaw scale beyond a single SQLite file?
Yes. ERPClaw is database-agnostic via PyPika, a SQL query-builder abstraction. SQLite is the default; PostgreSQL is a first-class alternative backend with no code changes required. For workloads that outgrow SQLite (high concurrent writes, very large datasets), PostgreSQL is the supported path.

### How do I migrate from ERPNext to ERPClaw?
ERPNext exports cleanly per DocType to CSV. The ERPClaw AI agent reads the exports and runs the matching import actions. Custom DocTypes get re-expressed as SKILL.md specs and ERPClaw modules; the AI assists that translation. Typical elapsed time is 4-8 weeks including a parallel-run period.

## One install, one database, every module
Free forever. Self-host in 5 minutes. AI-native by design.
Try the demoView on GitHub

Related: the parent ERPNext comparison, the migration guide, or the ERPNext AI ecosystem deep dive.

---

## Microsoft Dynamics Alternative: Free AI-Native ERP
URL: https://www.erpclaw.ai/compare/microsoft-dynamics/

# ERPClaw vs Microsoft Dynamics
Dynamics 365 is per-app licensed; the cost compounds fast. ERPClaw is open-source AI-native, every module included, $0 forever.
Same primitives. AI-native by design. Free.

| Feature | Microsoft Dynamics | ERPClaw
| Pricing | Business Central $80/user/mo Essentials, $110 Premium (Oct 2025 raise); Finance & Ops $210+ | $0 forever
| Implementation | $50K-500K + Microsoft partner | $0 (5-min install)
| Implementation time | 3-12 months | 5 minutes
| Modules | Per-app licensing (Sales, Service, Finance separate) | All 46 included always
| Multi-entity | Yes (F&O tier) | Yes
| AI capabilities | Copilot (chat sidebar on Dynamics) | AI-native architecture
| Customization | AL/Power Platform (Microsoft stack) | Spec-first; any LLM regenerates code
| Self-hosted | Mostly cloud; on-prem on legacy SKUs | Yes (SQLite or PostgreSQL)
| Open source | Proprietary | Free &amp; Open Source
| Vendor lock-in | Significant (Microsoft ecosystem) | None (cp data.sqlite)

## Key differences

### Per-app vs all-included
Microsoft Dynamics 365 splits Sales, Service, Finance, Supply Chain, HR into separate licensed apps; the cost compounds fast for a 10-person team. ERPClaw includes every module in the open-source build.

### Copilot is AI-decorated
Microsoft Copilot is a chat sidebar on top of Dynamics 365's existing data model. The AI does not autonomously post journal entries the way ERPClaw's AI-native architecture does. The 5-trait test on /ai-native-erp/ scores both layer by layer.

### AL framework vs spec-first regen
Dynamics 365 customizations live in AL (Microsoft's proprietary language) and Power Platform. ERPClaw customizations live in SKILL.md specs that any LLM regenerates from. No Microsoft partner required.

### Self-host without the lock-in
Even the on-prem Dynamics SKUs are tied to the Microsoft ecosystem (Azure, Power Platform, Microsoft 365 SSO). ERPClaw is Free &amp; Open Source, runs on any Linux/macOS/Windows machine, and uses portable SQLite or PostgreSQL.

## Frequently asked questions

### What's the difference between Dynamics 365 Business Central and ERPClaw?
Business Central is mid-market cloud ERP at $80-$110 per user per month (raised October 2025; first price increase in five-plus years) plus a Microsoft partner implementation. ERPClaw is Free &amp; Open Source at $0 forever, every module included, no per-user fee. The architectural difference is bigger than the price difference: ERPClaw is AI-native; Business Central is AI-decorated with Copilot on top.

### Is Copilot the same as AI-native?
No. Copilot is a chat sidebar on top of Dynamics 365's existing forms-and-workflows data model. The AI suggests; a human still drives every meaningful write. ERPClaw is AI-native: the agent posts journal entries autonomously across every module with constitutional invariants enforcing integrity.

### How does ERPClaw compare to Dynamics 365 Finance & Operations?
F&O targets large enterprise with deep multi-entity, multi-currency, and global-scale workflows. ERPClaw covers SMB to mid-market with the same primitives (multi-entity, 7 currencies, immutable GL) at $0. We do not compete at Fortune 500 scale; we replace the F&O entry tier where the pricing gets brutal.

### Can I migrate from Dynamics 365 to ERPClaw?
Yes. Dynamics exports cleanly to CSV per entity. The ERPClaw AI agent reads the export and runs the import actions to recreate your data. AL customizations need to be re-expressed as SKILL.md specs; the AI helps that translation. Plan 4-8 weeks elapsed including parallel run.

### When would I still pick Microsoft Dynamics?
If your stack is already deeply Microsoft (Azure, Power Platform, Microsoft 365 SSO, Teams), the ecosystem integration is a real benefit. If you have an existing Dynamics implementation and migrating is too disruptive, staying is reasonable. For most new mid-market deployments, ERPClaw covers the same primitives without the per-user license treadmill.

---

## NetSuite Alternative That Costs $0 Per User
URL: https://www.erpclaw.ai/compare/netsuite/

# NetSuite alternative: ERPClaw is open-source, AI-native, and free
NetSuite by Oracle is the mid-market and enterprise default at around $10K+/year plus $50K to $500K implementation, and its first-party finance AI arrives as read-only skills for your assistant. ERPClaw is open source, self-hosted (Docker, SQLite or PostgreSQL), AI-native from line one, $0 forever. Honest head-to-head, written by a co-founder.
By Nikhil Jathar, Co-founder, ERPClaw · Updated 2026-08-16

Looking for a NetSuite alternative? You probably already know the all-in cost: license from around $10K per year, implementation typically $50K to $500K, one to two full-time admins, plus renewal increases of 30 to 40 percent that are widely reported. The question is whether a structurally different product can cover the same primitives at a fraction of the cost.
NetSuite by Oracle (netsuite.com) is a real, well-built ERP with 30,000+ customers. ERPClaw is the open-source AI-native alternative. They differ on four things that actually matter for your decision.

- Architecture. NetSuite was designed in the late 1990s, and Oracle's finance AI reaches it from outside, through a connector, reading by default. ERPClaw is AI-native: the action layer is the AI's API, so the same assistant can answer and post.
- Delivery. NetSuite is Oracle cloud only. ERPClaw is self-host (Docker, SQLite or PostgreSQL); your data on your hardware.
- Economics. NetSuite is per-seat plus per-module gating plus implementation plus admin headcount. ERPClaw is $0 forever, every module included, 5-minute install.
- Lock-in. NetSuite migration off is famously expensive because of proprietary SuiteScript customizations. ERPClaw export is `cp data.sqlite somewhere-else.sqlite`.
This page is the honest head-to-head. We'll walk through when to pick NetSuite, when to pick ERPClaw, and where each falls short of the other. If you want the broader category map, start with our AI-native ERP framework.

## TL;DR comparison
The short version. Eight rows, side by side. The full reasoning is in the sections below. Prefer to take it with you? The ERPClaw vs NetSuite vs Rillet comparison PDF carries the same matrix in printable form.
| Dimension | NetSuite | ERPClaw
| Architecture | ERP designed in the late 1990s; Oracle's finance AI arrives as skills for your assistant, read-only by default | AI-native from line one; the action layer is the primary surface, and it can post
| Delivery | Oracle cloud only | Self-host (Docker, SQLite or PostgreSQL); your hardware, your data
| Pricing | License from around $10,000 per year, plus implementation in the $50K to $500K range, plus admin headcount | $0 forever, unlimited seats, every module included
| ERP scope | Full ERP plus SuiteApp ecosystem (deep but tier-gated) | Full ERP plus 14 industry verticals, every module included always
| Multi-currency | Yes (mature multi-book, ASC 830 FX) | 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED); invoice currency equals payment currency, no FX guessing
| Vendor lock-in | Significant (SuiteScript, proprietary schema, multi-year contracts) | None; export is `cp data.sqlite somewhere-else.sqlite`
| Open source | Proprietary, closed source | Free &amp; Open Source, full source on GitHub
| Best for | Public companies and large multi-entity operators who need vendor SOC 2 attestation | Solo founders to mid-market who want AI-native scope without per-seat lock-in

If you operate dozens of legal entities with monthly consolidation and need vendor SOC 2 attestation, NetSuite is the safer choice. If you want AI-native scope across every business function with no per-seat tax, open source, and your data on your hardware, that's ERPClaw.
NetSuite AI facts verified 2026-08-16 from Oracle's own NetSuite documentation and the published SuiteCloud agent-skills release.

## When NetSuite is the right choice
Credit where due: NetSuite is a real ERP with 30,000+ customers, deep multi-entity tooling, and a SuiteApp ecosystem most open-source alternatives can't match. Here are three situations where it beats us today.

### You operate many legal entities with intercompany consolidations.
NetSuite's multi-book accounting, intercompany elimination, and ASC 830 FX gain/loss handling at scale is more mature than ERPClaw's today. If your finance team closes for 20+ entities in 5+ currencies with monthly consolidation, NetSuite's polish on that specific workflow has real value. ERPClaw covers the primitives, but the operator UX for that profile is not where NetSuite's is.

### You need a SuiteApp ecosystem.
NetSuite's SuiteApp marketplace has thousands of pre-built vertical and integration apps (industry packs, tax engines, EDI, advanced revenue management). For organizations that pay for that breadth of pre-built integrations, the ecosystem is a real advantage. ERPClaw covers Stripe (live on Stripe Marketplace), Shopify (v1.1.3 OAuth Token Exchange), and 14 industry verticals natively, but doesn't yet have a marketplace of third-party SuiteApp-equivalents.

### Your buyer requires vendor SOC 2 and a contractual SLA.
Oracle is publicly attested under SOC 2 and SOX, with dedicated account managers and contractual support SLAs. If your CFO or audit committee mandates a vendor-managed SaaS with a named accountable counterparty, NetSuite's commercial structure fits that mandate directly. ERPClaw is open source plus the co-founder team; for some governance frameworks, that mismatches the requirement.

Staying on NetSuite for now? You do not have to wait for Oracle's AI roadmap: AI for NetSuite shows how to put a plain-language AI analyst beside the NetSuite books you already run.

## When ERPClaw is the right choice
Five situations where the structural choice points to us. Source on github.com/avansaber/erpclaw if you want to read before you decide.

### You want AI-native, not AI-decorated.
Oracle now ships real first-party finance AI for NetSuite, and it is worth understanding exactly what shape it takes: downloadable skills that install into an assistant you already use, reaching your data through a connector, read-only unless you authorise a change. That is genuine and useful. It is also analysis sitting beside a late-1990s core. ERPClaw's AI is the architecture instead: every action, from customer to invoice to payment to payroll, is invokable in plain language, and a checking engine verifies the books after every posting. Reading is where theirs stops by default and where ours starts.

### You want $0 forever, not $10K+ per year plus implementation.
NetSuite licenses run from around $10,000 per year for a small package and scale fast with users, modules, and revenue tiers; implementation typically lands at $50K to $500K, plus one to two full-time admins, plus renewal increases of 30 to 40 percent (widely reported). ERPClaw is $0 forever, every module included, install in 5 minutes. For a $10M to $100M revenue company, that math compounds.

### You want to own your data and your code.
NetSuite stores your books on Oracle's cloud. ERPClaw runs on your hardware (laptop, server, anywhere) in SQLite or PostgreSQL with documented schemas. Migration off NetSuite is famously expensive because of proprietary SuiteScript customizations and undocumented data structures. With ERPClaw, your data is portable from day one because the source is yours.

### You want full ERP without per-module tier gates.
NetSuite's pricing tiers gate access to revenue management, multi-entity, fixed assets, and other modules; each higher tier adds the next set. ERPClaw ships every module always: accounting, inventory, manufacturing, payroll (US: W-2, 1099, NACHA, FICA, FUTA, SUTA), Stripe, Shopify, plus 14 industry verticals. One install, one license, no upsell.

### You want any business action invokable from natural language.
ERPClaw's AI-native architecture means the action layer is the API. Type "add Bob from BigCo as a customer for 5 widgets at $50" and the AI invokes add-customer, add-sales-order, and the GL posting in one transaction with a full audit row. NetSuite's AI agents are scoped to specific finance and supply-chain workflows; ERPClaw's chat is the primary interface for the whole business.

For the deeper open-source argument, see open-source AI accounting. For the AI-native framework, see AI-native ERP.

## Pricing
Concrete cost items over a real horizon. NetSuite's exact quote depends on tier, modules, user count, and revenue band; the figures below are widely reported public ranges, not a binding quote.
| Cost item | NetSuite | ERPClaw
| License (per year) | From around $10,000, scales with users, modules, and revenue tier | $0 forever, unlimited seats
| Implementation | $50K to $500K, 6 to 18 months | 5-minute install, no implementation engagement
| Modules | Tier-gated; advanced revenue, multi-entity, fixed assets each become available at higher tiers | Every module included always
| Admin headcount | 1 to 2 full-time NetSuite admins typical | No dedicated admin required
| Renewal | 30 to 40 percent increases widely reported | No renewal; Free &amp; Open Source license is permanent
| Source code access | Closed | Free &amp; Open Source: fork it, audit it, run it
| Where it runs | Oracle cloud (your data on Oracle's servers) | Your hardware (laptop, server, anywhere)

For a $10M to $100M revenue mid-market company, the multi-year cost differential between NetSuite and ERPClaw is significant: NetSuite typically runs to six or seven figures across license, implementation, and admin headcount over three years; ERPClaw is zero. See our pricing page for the full breakdown.

## Migration: switching to ERPClaw
Switching ERPs is friction. Here is the honest path.
From NetSuite. Export your chart of accounts, customers, vendors, items, and transaction history via NetSuite's CSV / saved-search exports. Import via ERPClaw's `import-chart-of-accounts` and `import-opening-balances` actions. The friction is real: SuiteScript customizations don't port (you re-implement business logic as ERPClaw actions), and some custom record types may require schema additions.
Stripe history. ERPClaw's Stripe integration backfills directly from the Stripe API. You don't bring Stripe data through NetSuite first. Connect, run the backfill action, the GL entries post with full audit rows.
Shopify history. Same model. The Shopify integration (OAuth Token Exchange via App Bridge) pulls orders, payouts, and adjustments straight from Shopify.
Honest gap. A multi-entity NetSuite tenant with 5+ subsidiaries, multi-book, and ASC 830 FX is not a weekend port. Plan for a phased migration: one entity at a time, parallel-run the books for a quarter, then cut over. Reach us through support if you want a sized estimate, and see the product working at /demo/.

## What NetSuite does better than ERPClaw today
NetSuite wins on these dimensions today, and pretending otherwise would be insulting to a serious peer.

- Multi-entity consolidation depth at scale. Intercompany eliminations, multi-book accounting, ASC 830 FX gain/loss across many entities is more mature than ERPClaw's today.
- SuiteApp ecosystem breadth. Thousands of pre-built vertical apps, tax engines, EDI connectors, and advanced revenue modules. ERPClaw's third-party ecosystem is younger.
- Vendor accountability. Oracle is publicly SOC 2 / SOX attested with dedicated account managers and contractual SLAs. ERPClaw is Free &amp; Open Source plus the co-founder team; different accountability model.
- Polished web UI. ERPClaw is CLI plus chat first; the webclaw web dashboard exists but is limited for Stripe and Shopify today.

## What ERPClaw does that NetSuite doesn't
The structural wins. These are not feature checkmarks; they are architectural choices NetSuite cannot retrofit without rebuilding on top of a 1998 codebase.

- AI-native architecture. The action layer is the AI's API, so the assistant can both answer and post. Oracle's NetSuite finance skills are read-only by default and sit beside a late-1990s core.
- Free &amp; Open Source. Fork, contribute, audit. NetSuite is proprietary, closed source.
- Self-host. Your data on your hardware. NetSuite is Oracle cloud only.
- $0 forever, every module included. No per-module gating, no per-seat tax, no renewal increases.
- 5-minute install. NetSuite implementations run 6 to 18 months.
- No vendor lock-in. Export is `cp data.sqlite somewhere-else.sqlite`. NetSuite migration off is famously expensive.
- Multi-currency in 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED). Invoice currency equals payment currency, no FX guessing.
- 12-step GL invariant validation on every posting. Bad entries cannot reach the books.
- Stripe integration live on Stripe Marketplace. Shopify integration shipped at v1.1.3 (OAuth Token Exchange via App Bridge).
- Patent pending plus trademark filed. Long-term durability signal.

## Frequently asked questions

### How much does NetSuite actually cost all in?
Public NetSuite license starts around $10,000 per year for a small package and scales fast with users, modules, and revenue tiers. Implementation typically lands at $50,000 to $500,000 depending on scope, and ongoing administration commonly requires one to two full-time admins. Renewal increases of 30 to 40 percent are widely reported. ERPClaw is $0 forever, every module included, with a 5-minute install.

### Can I migrate off NetSuite without losing data?
NetSuite migration is famously expensive because of proprietary SuiteScript customizations and undocumented data structures. ERPClaw stores everything in SQLite or PostgreSQL with documented schemas; your data is portable from day one. The export pattern is literally `cp data.sqlite somewhere-else.sqlite`.

### Is ERPClaw really comparable to NetSuite at the mid-market?
ERPClaw covers the same primitives NetSuite does at the mid-market entry point: accounting, inventory, manufacturing, HR, payroll, CRM, projects, billing, multi-entity, multi-currency. We don't claim Fortune 500 parity. We do claim the structural moats matter more for a $10M to $100M revenue company: open source, self-hosted, $0 forever, AI-native architecture NetSuite cannot retrofit onto a 1998 codebase.

### Does NetSuite have AI in 2026?
Yes, and more than it did a few months ago. Alongside the Oracle AI agents applied to NetSuite workflows, Oracle now publishes first-party finance skills for NetSuite itself: you install them into an assistant you already use, they reach your data through Oracle's NetSuite connector, and they cover statements, variance, reconciliation review, close guidance, aging and executive narratives. They default to read-only unless you explicitly authorise a change. So the honest comparison is no longer whether NetSuite has AI. It is that theirs analyses and ours also acts: in ERPClaw the AI is the primary user of the action layer, and a checking engine verifies the books after every posting.

### Why is NetSuite called AI-decorated and not AI-native?
No, and Oracle's own delivery shape shows why. Its NetSuite finance AI arrives as skills you load into a separate assistant, talking to NetSuite through a connector, rather than as the way the product itself works. The data model, workflow engine and integration patterns still predate the AI era. That is a reasonable way to add AI to a system designed in the late 1990s; it is not the same as being built for it. The five-trait test on our AI-native ERP pillar walks the difference layer by layer.

### What does ERPClaw not do that NetSuite does?
ERPClaw's web dashboard is included free and currently in beta; most users work through the chat interface, which stays the fastest path today. Multi-entity consolidation with intercompany eliminations is built in, and eliminations are safe to re-run; currency translation is recorded manually rather than computed, and transactional ASC 830 FX revaluation is intentionally out of scope, not a roadmap commitment. 24/7 vendor-managed support with a contractual SLA is not available; support is community-tier through GitHub plus the co-founder team. SuiteApp-equivalent third-party marketplace breadth is not yet there.

### What if my CFO insists on vendor SOC 2 attestation?
Honest answer: ERPClaw is not yet SOC 2 attested as a vendor (we are open-source software you self-host, not a SaaS counterparty). If your CFO or audit committee requires a contracted SOC 2 SaaS vendor with a named SLA, NetSuite or Sage Intacct fit that mandate directly. If your CFO is comfortable with open source plus self-host (the same way they accept Linux on production servers), ERPClaw's chain-hashed audit trail and full invariant validation pipeline give the controls auditors look for.

## Where to go next
Free forever. Install in 5 minutes. Every module included.
Talk to a co-founder Read the AI-native framework
Or browse the source on github.com/avansaber/erpclaw.

---

## ERPClaw vs Odoo: Features, AI, Pricing, and Tradeoffs
URL: https://www.erpclaw.ai/compare/odoo/

# ERPClaw vs Odoo
Odoo is form-based. ERPClaw is chat-based.
Same open-source ethos. AI-native rebuild. SQLite simplicity.

## Why teams look for an Odoo alternative
Odoo is one of the most popular open-source ERPs in the world, so most people searching for an Odoo alternative are not trying to leave open source. They are running into the gap between what Odoo Community gives away and what Odoo actually charges for once a business grows.
The first wall is the Community versus Enterprise split. Odoo Community is free to license, but Odoo Studio, the mobile app, and many of the apps teams reach for daily sit behind the paid plans. To unlock them you move to Standard at $31.10 per user per month or Custom at $61 per user per month (odoo.com/pricing, annual billing). A ten-person company on Custom pays $7,320 in year one before anyone touches the system. The Community vs Enterprise breakdown maps exactly which features, and which AI, sit on each side of that line.
The second wall is implementation. Odoo's customization model assumes a partner. Forum-reported engagements run from $15,000 for 170 hours to $20,000 with the project still unfinished a year later, and every later change is a fresh quote. The license is open; the operating cost is not.
The third wall is AI. Odoo 19 ships genuine native AI (Predictive Lead Scoring, AI Fields, Semantic Search, and RAG agents), but those features require an Enterprise plan. Community users get none of it. An Odoo alternative worth switching to has to answer all three: no tier gating, no partner tax, and AI built into the core rather than sold as an upgrade. That is the comparison below.

Comparison basis: US Standard and Custom pricing with annual billing, captured June 2026. For the detailed cost math and AI entitlement snapshot, read the Odoo Enterprise AI pricing analysis.
| Feature | Odoo | ERPClaw
| Price (community) | Free | Free
| Price (enterprise) | $31.10/user/month (Standard), $61 (Custom) | $0 self-hosted; Cloud Managed by request
| Architecture | Monolithic (Python + JS) | Modular skills (Python)
| Database | PostgreSQL required | Database-agnostic via PyPika; SQLite default, PostgreSQL fully supported
| Primary interface | Web forms | Natural language chat
| AI integration | Odoo 18/19 native AI (Lead Scoring, Document OCR, AI Fields, Semantic Search, RAG agents); OdooBot proxies ChatGPT (OpenAI) | AI-native from line one; every action invokable by an agent
| AI tier gating | Native AI requires Odoo Enterprise (Standard $31.10/user/month, annual); Community has zero AI | Same AI in every install; no Community vs Enterprise split
| Foundation model | OdooBot depends on ChatGPT (OpenAI, third-party) | Model-agnostic; works with any LLM the user chooses
| Setup time | Days to weeks | 5 minutes
| Configuration | Heavy ERP customization | Skills + chat = no config screens
| Module ecosystem | 13,000+ third-party apps | the whole ERP plus 14 verticals
| Community | Massive | Growing
| Industry depth | Many vertical packages | 5 native verticals (health, edu, etc.) + 13 more
| Self-hosted | Yes | Yes

Odoo AI feature facts verified 2026-05 from Odoo's public documentation (odoo.com/documentation/19.0) and Braincuber Q1 2026 review.

## Key differences

### Different paradigm
Odoo is forms-and-fields. ERPClaw is chat. The AI-native architecture is not a feature on top of an ERP; it's the architecture itself.

### Simpler footprint
Odoo requires PostgreSQL and extensive setup. ERPClaw is database-agnostic via PyPika: SQLite is the default for fast install and one-file backups, PostgreSQL is fully supported for teams that prefer it. Same code, your choice.

### Specifications instead of customization
Odoo customization means writing Python modules and tweaking forms. ERPClaw extension is updating SKILL.md specs and regenerating.

### Younger but moving fast
Odoo has 15 years and 13,000 third-party apps. ERPClaw has 8 weeks and one integrated system. We're not pretending parity yet; we're betting on a different architecture.

## Why switch now?
Three things changed in 2026 that make this the right moment to move.

### Odoo 19's AI is Enterprise-only
The 19.0 release put real AI into the product (Semantic Search, AI Fields, RAG agents), but it lives behind the paid Enterprise plans. If you run Community, the 2026 AI story does not include you. ERPClaw ships the same AI action layer in every install, with no Community versus Enterprise line and no per-seat gate.

### PostgreSQL is a first-class ERPClaw backend
The most common reason teams stayed on Odoo was its PostgreSQL requirement. ERPClaw runs on SQLite by default and supports PostgreSQL as a fully supported option through its PyPika query layer, so the same code targets either. Moving off Odoo no longer means giving up the database you standardized on.

### The Stripe integration is live
ERPClaw's Stripe addon is published on the Stripe Marketplace and handles real revenue flows, including ASC 606 deferred revenue and net-of-fees payout decomposition. Payments are shipping today, not a roadmap promise.

---

## Odoo Community vs Enterprise: What AI You Get
URL: https://www.erpclaw.ai/compare/odoo/community-vs-enterprise/

# Odoo Community vs Enterprise: What AI You Actually Get
Choosing between Odoo Community and Enterprise in 2026? The AI question changes the math.

## The split
Odoo Community is free and open source. In 2026, it ships with zero native AI features. OdooBot is in Community, but the ChatGPT-style integration is Enterprise.
Odoo Enterprise is $24.90 per user per month. Every native AI feature, current and roadmapped, sits behind that subscription.
Pricing basis used on this page: the published tier captured May 24, 2026. Treat it as the source basis for this edition comparison, not as a current quote. Local taxes, currency, plan selection, and partner discounts can vary.

## What's in Community (no native AI)
Community is a complete ERP. The gap is specifically AI.

### Included

- +CRM, accounting, inventory, manufacturing, HR, project, purchase, sales (full set)
- +Multi-company, multi-currency, multi-warehouse
- +OdooBot exists, but the ChatGPT-style integration is Enterprise-only
- +Solid reporting and dashboards
- +Free, GPL-style license, self-hostable

### Not in Community

- xPredictive Lead Scoring
- xAI Document OCR
- xAI Fields
- xAI App (sales forecasting, stock demand)
- xSemantic Search
- xRAG agents (v19)
- xAgentic AI (v20 roadmap)

## What's gated behind Enterprise ($24.90 per user per month)
The full 2026 native AI feature set, plus the agentic roadmap.

### Predictive Lead Scoring
Ranks incoming CRM leads by probability of conversion. Trained on your closed-won history.

### AI Document OCR
Pulls vendor bills, expense receipts, and bank statements into structured records.

### AI Fields
Auto-fills computed fields on records using an LLM call. Triggers on create or write.

### AI App for forecasting
Sales forecasting and stock demand prediction across the inventory and sales modules.

### Semantic Search
Vector-backed search across Odoo records. Replaces strict keyword match.

### RAG agents (v19)
Retrieval-augmented agents that answer questions against your Odoo instance.

### Agentic AI (v20 roadmap)
Announced for v20: agents that take actions inside Odoo. On the published roadmap, not shipped yet.

## Concrete math: what Enterprise costs over 1, 3, and 5 years
Subscription only. The AI features are bundled into this price; you cannot buy AI a la carte.
| Team size | 1 year | 3 years | 5 years
| 5 users | $1,494 | $4,482 | $7,470
| 10 users | $2,988 | $8,964 | $14,940
| 25 users | $7,470 | $22,410 | $37,350

Based on published $24.90 per user per month tier. Annual figures are 12 months at full price. Local taxes, currency, and partner discounts vary.

## The open-source alternative without tier gating
ERPClaw is open source (Free & Open Source), AI-native (the AI agent posts to GL directly), and $0 forever. Single install. Single license. No per-user fee. No separate AI subscription.
If you wanted Odoo Community's price but Odoo Enterprise's AI, ERPClaw is the shape of product that gives you both. Self-hosted on your own infrastructure, every module included. If you are planning a move, use the Odoo migration playbook for export, mapping, validation, and cutover.

## Frequently asked questions

### Does Odoo Community include AI in 2026?
No. Odoo Community is the free, open-source edition and does not include the native AI feature set. OdooBot exists in Community, but the ChatGPT-style integration, Predictive Lead Scoring, AI Fields, AI Document OCR, AI App, Semantic Search, and the RAG and agentic capabilities all sit inside Odoo Enterprise.

### What does Odoo Enterprise's $24.90 per user per month actually include?
Odoo Enterprise at $24.90 per user per month includes all native AI features (Predictive Lead Scoring, AI Document OCR, AI Fields, AI App for forecasting, Semantic Search, v19 RAG agents) plus the Enterprise-only modules (Studio, IoT, Quality, additional accounting localizations, Sign, Marketing Automation, and others). The published per-seat figure is the standard cloud tier; on-premise and self-hosted Enterprise pricing is comparable per user.

### Can I add AI to Odoo Community via third-party apps?
Yes, partially. There are community apps in the Odoo App Store that wrap third-party LLM APIs around Community records (typically OpenAI, Anthropic, or self-hosted Llama). These are not native to Odoo, they require separate maintenance, and they do not match the native AI App's depth on forecasting. The OEC.sh and Braincuber writeups cover the available options in detail.

### What if I want open source AND AI in the same product?
ERPClaw is open source (Free &amp; Open Source), AI-native by design (the AI agent posts to GL directly through the same action layer a human uses), and $0 forever. No tier gate, no per-user fee, no separate AI subscription. Self-hosted on your own infrastructure.

Sources: Odoo 19 AI documentation (odoo.com/documentation/19.0/applications/productivity/ai.html), OEC.sh AI for Odoo Community Edition guide, Braincuber Odoo AI Modules and Pricing writeup.

## Free, open source, AI-native
Install in 5 minutes. Every module. No tier gate.
Try the demoView on GitHub

Related: the parent Odoo comparison, the AI-native ERP framework, or the Odoo Enterprise AI pricing deep dive.

---

## Puzzle Alternative That's Free and Whole-Business
URL: https://www.erpclaw.ai/compare/puzzle/

# Puzzle alternative: whole business, your machine, $0
Puzzle (puzzle.io) is AI accounting for startups: real automation, published pricing, books only. ERPClaw is the open-source AI-native ERP: the books plus inventory, manufacturing, purchasing, payroll, and Sales, self-hosted, free forever.
By Varun Borawake, Co-founder, ERPClaw · Published 2026-08-14

Puzzle's headline is "Accurate Books. At AI Speed.", and the company behind it is real: founded by Sasha Orloff, a $15M Series A led by General Catalyst in February 2023 and a further $30M announced that November, $50M in total. The product is a genuine double-entry system with AI categorization claiming up to 98 percent automation, aimed at venture-backed startups and the accounting firms that serve them.
The comparison comes down to three questions:

- Scope. Puzzle keeps books and stops there: no inventory, no purchasing, no payroll, no tax filing, by their own site and FAQ. ERPClaw is the whole ERP in one install.
- Custody. Puzzle is SaaS only. ERPClaw lives on your hardware with open source code.
- Economics. Puzzle publishes tiers from $25 to $300 per month with AI metered in credits. ERPClaw is $0 forever with nothing metered.
For the wider category map, start with our AI-native ERP framework and the suite roll-up comparison.

## TL;DR comparison
| Dimension | Puzzle | ERPClaw
| Type | Commercial AI accounting for startups | Open-source AI-native ERP (whole business)
| Delivery | Vendor SaaS only | Self-host (Docker, CLI, SQLite or PostgreSQL)
| AI focus | Auto-categorization, AI reconciliations, close automation, metered by AI credits | Natural-language action layer across every business operation, no meter
| Pricing | Published: $25 to $300 per month by tier (annual billing) | $0 forever, open source
| Scope edges | Books only: no inventory, purchasing, payroll, or tax filing | Inventory, manufacturing, purchasing, payroll, and Sales built in
| Best for | Venture-backed startups on Stripe, Mercury, Brex who want managed books | Businesses that want the whole operation in one system they own

## When Puzzle is the right call

### You are a lean startup and want managed SaaS books this afternoon.
Puzzle's pitch is startup-shaped: connect Stripe, Mercury, Brex, Ramp, and get accurate books with burn and runway on a dashboard. If your whole operation is a SaaS product and a bank account, that focus is a feature. If the only thing keeping you off ERPClaw is not wanting to host it, Cloud Managed is our run-it-for-you option, by request through /pricing/.

### Their integration list is your exact stack.
Stripe, Mercury, Brex, Ramp, Gusto, Rippling, Deel: Puzzle leads with the tools venture-backed startups already use, and meets them with published, transparent pricing, which deserves credit in a category that hides its numbers.

### You want your accountant and the software from one ecosystem.
Puzzle markets to accounting firms running many startup clients, and pairs with partner firms for taxes. If you want software chosen by your bookkeeper rather than software you operate, that channel matters.

## When ERPClaw is the right call

### Your business is more than a bank feed and a Stripe account.
Puzzle keeps books. It does not do inventory, purchasing, payroll (it integrates Gusto and Rippling), or tax filing (partner firms), and its own site confirms the shape. ERPClaw runs stock with hard reservations, manufacturing, purchasing, payroll, and built-in Sales in the same database as the ledger.

### You want $0 with no meter, not a tier ladder.
Puzzle's published pricing is honest and starts at $25 per month, with AI usage metered in credits per plan. ERPClaw does not meter the interface to the product: every plain-language operation is included, forever, at $0, because open source is the pricing model.

### You want your books on your own machine.
Puzzle is SaaS only. ERPClaw is self-hosted by design, SQLite or PostgreSQL, Free &amp; Open Source, and the ledger is a file you can open and audit yourself. For founders who consider their books a crown jewel, custody is the comparison.

### You want AI-native to mean the whole company, not the close.
Both products are AI-native rather than AI-decorated, and Puzzle's categorization and reconciliation automation are real. The reach differs: ERPClaw's plain-language layer runs every operation, from 'invoice Acme for the retainer' to 'hold 40 chairs for this order,' with deterministic code posting and an invariant engine checking the books after every entry.

### You plan to outgrow startup accounting without a second migration.
Startups that make it become companies with warehouses, purchase orders, and employees in several entities. Puzzle's answer at that point is another migration. ERPClaw's answer is that the ERP was already there, free, waiting for you to grow into it.

## What it costs
| Item | Puzzle | ERPClaw
| Subscription | $25 / $60 / $100 / $300 per month by tier, billed annually (published, checked 2026-08-14) | $0 forever, unlimited users
| AI usage | Metered in AI credits per plan | Unmetered; the plain-language layer is the product
| Source code | Closed | Free &amp; Open Source: fork it, audit it
| Where it runs | Their cloud only | Your hardware, your choice of database

## Common questions

### Is Puzzle AI-native or AI-decorated?
AI-native for its category. Puzzle is a double-entry accounting system built around AI categorization, reconciliation, and close automation, with agent features shipping under its own brand. The honest comparison is scope and custody, not the authenticity of the AI.

### Doesn't Puzzle publishing prices make them the good guys?
It genuinely does them credit; most of this category hides pricing behind demos. Their tiers run $25 to $300 per month billed annually, with AI usage metered by credits. ERPClaw's price is also published: $0, with nothing metered.

### Can Puzzle run inventory or purchasing?
No. Nothing on their homepage, pricing page, or comparison pages mentions inventory, purchase orders, or warehouse capability (checked 2026-08-14). Puzzle is books for startups, and says so. ERPClaw ships the full operational stack in the same install as the books.

### Does Puzzle file my taxes or run payroll?
By their own FAQ, no on both: taxes go through partner firms, payroll through integrations like Gusto and Rippling. ERPClaw ships US payroll in the ERP itself; tax filing is your accountant's job in both worlds.

### Who should honestly pick Puzzle?
A venture-backed software startup with no physical operations, on Stripe and a modern bank, that wants managed SaaS books at a fair published price. That is a real fit. The moment the business grows things, people, or entities to run, the whole-business, self-hosted side of this table is ERPClaw's.

## Books plus the business they belong to
ERPClaw: the AI-native ERP. Run your whole business in plain language. Free forever, self-hosted, PostgreSQL or SQLite.
About ERPClawJust want answers? ERPClaw Insights

---

## QuickBooks Alternative - Free Open Source ERP
URL: https://www.erpclaw.ai/compare/quickbooks/

# ERPClaw vs QuickBooks
QuickBooks is great accounting software. ERPClaw is a complete ERP that also happens to be free, open source, and AI-native.

| Feature | ERPClaw | QuickBooks
| Price | Free (open source license) | $38 to $235 a month, per plan
| Users | Unlimited | Per-user pricing
| Interface | Natural language chat | Web forms and menus
| Setup time | 5 minutes | Days to weeks
| Training | None needed | Hours of tutorials
| Self-hosted | Yes (your server) | No (cloud only)
| Open source | Yes (open source) | No (proprietary)
| Data ownership | 100% yours | Intuit's cloud
| AI capabilities | AI-native: you run it by asking, not by menus | Intuit Assist sidebar (human approves every action)
| Database | Your data in a database you own; SQLite default, PostgreSQL fully supported | Closed proprietary cloud storage
| Inventory | Full (batches, serials, manufacturing) | Basic (Plus plan only)
| Manufacturing | Included (BOM, MRP, work orders) | Not available
| HR & Payroll | Included | Separate add-on ($$$)
| Multi-company | Included | Separate subscription per company
| API access | Full (open source) | Limited, paid tiers
| Offline access | Always (local database) | Limited

QuickBooks + Intuit Assist facts verified 2026-05 from Intuit's public product pages.

## Key Differences

### Full ERP, Not Just Accounting
QuickBooks handles accounting well. ERPClaw does accounting plus inventory, manufacturing, HR, CRM, projects, quality, and more. One system for everything.

### No Monthly Fees, Ever
QuickBooks charges from $38 a month to $235 a month depending on plan, with add-ons that stack on top (prices from Intuit's pricing page, checked July 2026). ERPClaw is open source and free to self-host with unlimited users.

### Chat vs. Click
QuickBooks requires navigating menus and filling forms. ERPClaw lets you do everything through natural conversation. "Create an invoice for Acme, $5,000 for consulting."

### Your Data, Your Server
QuickBooks stores your financial data on Intuit's servers. ERPClaw runs entirely on your infrastructure. No vendor has access to your books.

### Self-Improving ERP
QuickBooks cannot self-improve. ERPClaw OS detects what your business needs, generates new modules from patterns, and validates every change against constitutional articles. No other ERP does this.

## Who should consider switching?
You're paying for QuickBooks Plus or Advanced and need inventory/manufacturing

You're tired of per-user pricing as your team grows

You want AI that actually understands your business, not a basic chatbot

You need your data on your own servers for compliance or preference

You want one system instead of QuickBooks + separate HR + separate inventory tools

---

## Rillet Alternative That's Free and AI-Native
URL: https://www.erpclaw.ai/compare/rillet/

# Rillet alternative: ERPClaw is open-source, self-hosted, and free
Both are AI-native. Rillet is closed-source SaaS focused on finance close, with seat-based pricing. ERPClaw is open source, self-hosted (Docker, SQLite or PostgreSQL), full ERP scope, $0 forever. Honest head-to-head, written by a co-founder who shipped the integrations.
By Varun Borawake, Co-founder, ERPClaw · Updated 2026-08-16

Looking for a Rillet alternative? You probably aren't shopping for AI-decorated bookkeeping with a chat box on top. You've already decided on AI-native. The question is delivery model, scope, and economics.
Both Rillet, Inc. (rillet.com) and ERPClaw are AI-native by design. That wedge has been won. They differ on three things that actually matter for your decision.

- Delivery. Rillet is a commercial SaaS. ERPClaw is open source, self-hosted via Docker, CLI, Postgres, or SQLite.
- Scope. Rillet's AI is built around finance close: reconciliations, variance analysis, board-deck-ready financials. ERPClaw's AI is built around action: every business operation (customer, invoice, payroll, inventory, payment, tax) is invokable from a natural-language prompt.
- Economics. Rillet is seat-based pricing. Founder-reported quotes land in the $2,000 to $10,000 per month range, plus a 45-day implementation. ERPClaw is $0 per seat forever, install in 5 minutes.
This page is the honest head-to-head. We'll walk through when to pick Rillet, when to pick ERPClaw, and where each falls short of the other. ChatGPT, asked to compare them directly, classifies us this way:
"Rillet: Commercial AI-native ERP. Best for VC startups, SaaS teams. Hosting: Vendor SaaS. AI angle: Finance automation, close. ERPClaw: Open-source / self-hosted AI ERP. Best for builders wanting full control. Hosting: Self-hosted. AI angle: Natural-language actions."
Where Rillet stands as of August 2026, from their own site and dated press: "raised over $100M from Sequoia, a16z, and ICONIQ" and "500+ customers, including public companies with over $1B in ARR" on their about page (checked 2026-08-16), built from a $25M Series A (May 2025) and a $70M Series B (August 2025), partnerships with BILL (March 2026) and EY (April 2026), and pricing that remains demo-gated with no published figures. Worth noting their homepage still says 400+ finance teams while their about page says 500+, so we cite the higher figure they publish about themselves. Their integrations page lists payroll and procurement as integrations only, and no inventory coverage at all, which keeps the scope wedge below exactly where it was.
Both products are real. The right answer depends on which side of those dimensions you're on. If you want the broader category map (where Rillet, Numeric, Puzzle, Digits, and ERPClaw all sit), start with our AI-native ERP framework.

## TL;DR comparison
The short version. Six rows, side by side. The full reasoning is in the sections below.
| Dimension | Rillet | ERPClaw
| Type | Commercial AI-native ERP | Open-source AI-native ERP
| Delivery | Vendor SaaS | Self-host (Docker, CLI, cloud)
| AI focus | Finance close, reconciliations, variance analysis | Natural-language action layer (any action invokable from a prompt)
| Pricing | Seat-based, contact sales | $0 per seat forever, open source license
| Best for | VC startups, SaaS finance teams who want SaaS-managed close | Builders and technically-aware finance teams who want control plus scope
| Honest gap vs the other | No self-host, no source code, scope is finance | CLI-first today; the close runs inside the full ERP rather than a dedicated cockpit

If you want a vendor-managed, polished SaaS focused purely on finance close, with a contractual support relationship, Rillet is a strong choice, and ERPClaw Cloud Managed (by request via /pricing/) covers the vendor-run preference with ERPClaw underneath. If you want full ERP scope (customer, invoice, inventory, payroll, payment, tax) with AI-native action across all of them, plus open source and self-host, that's ERPClaw.

## When Rillet is the right choice
Credit where due: Rillet is a real AI-native finance product, well-funded, with paying customers. Here are five situations where it beats us today.

### You want SaaS, not self-host.
Rillet is vendor-managed: updates push automatically, nothing to install. That delivery model has real value, and it is no longer a reason to leave this table, because ERPClaw Cloud Managed is the same idea with ERPClaw underneath: we run it for you, onboarding by request through /pricing/. Where Rillet still wins this point is the contractual support relationship a larger buyer may require.

### Finance close is your top pain.
Rillet's close workflow (reconciliations, variance analysis, anomaly flagging, board-ready financials) is polished, and it is built around that one ritual. ERPClaw closes the books too, inside the ERP that runs everything else, with automated close set up per business. If a purpose-built cockpit UI for the close team is the deciding factor, Rillet's is the more refined screen today.

### You're a VC-backed startup with SaaS finance ops.
Rillet was designed for this profile. Fast close, board-deck-ready financials, the integrations a Series A-to-C SaaS company expects (Stripe, Brex, Ramp, Salesforce). The product fit is direct. You will not have to bend the tool to your shape.

### You don't need full ERP scope.
If you only need GL, AP, AR, and close, Rillet is sufficient. ERPClaw covers inventory, payroll, manufacturing, support, projects, fleet, and 14 industry verticals. For a 30-person SaaS company with no warehouse and no factory floor, most of that is overkill.

### You want a vendor's public roadmap and release schedule.
Closed-source SaaS comes with a vendor that ships updates on a public schedule. ERPClaw is community plus a co-founder team. Different accountability model. Some buyers prefer the vendor contract.

## When ERPClaw is the right choice
Five situations where the structural choice points to us. Source on github.com/avansaber/erpclaw if you want to read before you decide.

### You want open source.
open source license. Full source on GitHub. Fork it, contribute, audit every line. Rillet is closed by design. If reading the GL posting code matters to you (or to your auditor), that's a binary choice.

### You want to self-host.
Your data on your machine. Postgres or SQLite. No vendor pulling SaaS subscriptions, no data residency review, no "what happens if they get acquired." If a customer asks where their financials live, the answer is your hardware.

### You want full ERP, not finance-only.
ERPClaw covers AR, AP, GL, payroll (US: W-2, 1099, NACHA, FICA, FUTA, SUTA), inventory, tax, Stripe (live on Stripe Marketplace), Shopify (OAuth Token Exchange). Rillet's scope is finance close. Different products. If you need to invoice, ship, pay employees, and reconcile in one shared database, that scope difference is the whole conversation.

### You want any action invokable from natural language.
ERPClaw's AI-native architecture means the action layer is the API. Type "add Bob from BigCo as a customer for 5 widgets at $50" and the AI invokes add-customer, add-sales-order, and the GL posting in one transaction with a full audit row. Rillet has chat for finance questions. ERPClaw's chat is the primary interface for the whole business.

### You want $0 forever.
No seat tax. No usage tier. No upsell. Every action across every module ships as open source.

For the deeper open-source argument, see open-source AI accounting.

## Pricing
What it really costs over a real horizon. Rillet's seat pricing is not on their public site (checked 2026-08-14); they route prospects through sales for a quote, so the range below is founder-reported, and your own quote is the number that matters.
| Cost item | Rillet | ERPClaw
| Subscription (per seat) | Seat-based, contact sales (founder-reported quotes: $2,000 to $10,000 per month) | $0 forever, unlimited seats
| Implementation | 45-day engagement, often $20K to $50K | Install in 5 minutes
| Source code access | Closed | Free &amp; Open Source: fork it, audit it, run it
| Where it runs | Vendor cloud (your data on their servers) | Your hardware (laptop, server, anywhere)

The math is rough by necessity: your exact Rillet quote will depend on tier, integrations, and contract length, so treat the table as the shape, not the invoice. The ERPClaw self-host costs are estimates for a small team running on a single VPS. Total cost differential is meaningful at scale; less meaningful for a one-person finance team where vendor management saves time worth more than the seat fee.
For a 1-person finance team at a small startup, the time-saving of Rillet's vendor management may exceed the cost differential. For a 5+ person team or a longer horizon, ERPClaw's cost advantage compounds. See our pricing page for the full breakdown.

## Migration: switching to ERPClaw
Switching ERPs is friction. Here is the honest path.
From Rillet. Export your chart of accounts and transactions as CSV. Import via ERPClaw's `import-chart-of-accounts` and `import-opening-balances` actions. Rillet supports CSV export, so the path is mechanical.
From QuickBooks Online or Xero. Same CSV path. ERPClaw's import actions accept the standard chart-of-accounts and trial-balance shapes. There is no one-click migrator yet; we are honest about that.
Stripe history. ERPClaw's Stripe integration backfills directly from the Stripe API. You don't bring Stripe data through Rillet first. Connect, run the backfill action, the GL entries post with full audit rows.
Shopify history. Same model. The Shopify integration (OAuth Token Exchange via App Bridge) pulls orders, payouts, and adjustments straight from Shopify.
Bank history. A Plaid connector links your bank accounts and syncs transactions, and the statement importer reads OFX, CAMT.053, MT940, and BAI2 with auto-matching against open invoices and payments. Both paths land in the same ledger, so moving off Rillet does not mean losing the bank connection. See importing a bank statement.

## What Rillet does better than ERPClaw today
Rillet wins on these dimensions today, and pretending otherwise would be insulting to a serious peer.

- Polished web UI. ERPClaw is CLI plus chat first; the web dashboard exists but is limited for Stripe and Shopify today.
- Dedicated close cockpit (reconciliations, variance analysis, anomaly flagging) as a purpose-built screen. ERPClaw closes the books inside the ERP that runs everything else; the cockpit polish is the difference, not the capability.
- Native multi-entity polish at scale. ERPClaw has intercompany and consolidation actions, but Rillet's UX for multi-entity operators is more refined.
- Vendor support tiers. ERPClaw is GitHub issues plus the co-founder team; Rillet has paid support contracts.

## What ERPClaw does that Rillet doesn't
The structural wins. These are not feature checkmarks; they are architectural choices Rillet cannot retrofit without rebuilding.

- Free &amp; Open Source. Fork, contribute, audit. Rillet is closed.
- Self-host. Your data on your hardware. Rillet is vendor cloud.
- Full ERP scope covering every major business function across 14 industry verticals. Rillet's scope is finance close.
- Action layer as API. Any action invokable from prompt or programmatic call.
- Stripe integration live on Stripe Marketplace. Shopify integration shipped at v1.1.3 (OAuth Token Exchange via App Bridge).
- Multi-currency in 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED). Invoice currency equals payment currency, no FX guessing.
- 12-step GL invariant validation on every posting. Bad entries cannot reach the books.
- Patent pending plus trademark filed. Long-term durability signal.
- $0 per seat forever.

## Frequently asked questions

### Is ERPClaw really free, or is there a paid tier coming?
Free forever, Free &amp; Open Source. There is no paid tier and no plan to introduce one. Every action across every module ships open. Cloud Managed, where we run it for you, is available by request through /pricing/, and the open-source product stays open and free regardless. The license is the contract.

### Can ERPClaw replace Rillet entirely?
For most use cases, yes. Both cover AR, AP, GL, and close. ERPClaw also covers inventory, payroll, tax, and Stripe and Shopify integrations. The honest gap: Rillet's close-specific workflow UI is more polished today. If your team specifically values that polish, evaluate side by side. For everything else (scope, source code access, self-host, $0), ERPClaw is the cleaner answer.

### What about my Stripe and Shopify history?
ERPClaw's Stripe integration backfills directly from the Stripe API. The Shopify integration (OAuth Token Exchange) does the same. You don't need to bring transaction data through Rillet first. Connect the source, run the backfill action, and the GL entries post with full audit rows.

### How long does migration take?
Rough estimate for a 5-person finance team migrating 1 year of history: 2 to 3 days. CSV exports from Rillet or QuickBooks, then ERPClaw's import-chart-of-accounts and import-opening-balances actions, then reconciliation. Stripe and Shopify connectors handle live transaction data without a re-export step. The friction is real but bounded.

### What if I need vendor support, not community?
Co-founders are accessible: reach us through /support/ and a founder answers, not a ticket queue robot. If you want SaaS-style operations with ERPClaw underneath, Cloud Managed is our hosted offering, onboarded by request through /pricing/. The open-source product stays free regardless. If a contractual vendor SLA is non-negotiable for your buyer, Rillet offers one today.

## Where to go next
Free forever. Install in 60 seconds. Every module included.
Talk to a co-founder Read the AI-native framework
Or browse the source on github.com/avansaber/erpclaw.

## Sources referenced in this article

- ChatGPT GPT-5 conversation captured 2026-05-05.
- Rillet public site (rillet.com) reviewed 2026-05-05; pricing not public, sales-quoted.
- ERPClaw `module_registry.json`: full action catalog across every module, signed and version-pinned.
- Stripe Marketplace listing for ERPClaw.
- github.com/avansaber/erpclaw (open source license).

---

## Sage Intacct Alternative That's Free and AI-Native
URL: https://www.erpclaw.ai/compare/sage-intacct/

# Sage Intacct alternative: ERPClaw is AI-native, free, and open source
Sage Intacct by Sage is the mid-market finance default at around $15K+/year plus add-on modules priced higher than the core, with Sage Copilot added on top. ERPClaw is open source, self-hosted (Docker, SQLite or PostgreSQL), AI-native from line one, $0 forever. Honest head-to-head, written by a co-founder.
By Varun Borawake, Co-founder, ERPClaw · Updated 2026-05-29

Looking for a Sage Intacct alternative? You probably already know the all-in cost: core license from around $15K per year plus per-seat scaling, add-on modules (fixed assets, subscription billing, multi-entity consolidation, project accounting) commonly priced higher than the core, plus implementation in the $10K to $100K+ range. The question is whether a structurally different product can cover the same primitives without the add-on tax.
Sage Intacct by Sage (sage.com/intacct) is a real, well-built mid-market finance platform with audit-firm familiarity and AICPA preferred-provider status. ERPClaw is the open-source AI-native alternative. They differ on four things that actually matter for your decision.

- Architecture. Sage Intacct is a 1990s general ledger product; Sage Copilot is a paid chatbot module added on top. ERPClaw is AI-native: the action layer is the AI's API.
- Delivery. Sage Intacct is Sage cloud only. ERPClaw is self-host (Docker, SQLite or PostgreSQL); your data on your hardware.
- Economics. Sage Intacct is per-seat plus per-module add-on plus implementation. ERPClaw is $0 forever, every module included, 5-minute install.
- Scope. Sage Intacct's scope is mid-market finance plus add-ons. ERPClaw is full ERP (inventory, manufacturing, payroll, Stripe, Shopify) plus 14 industry verticals in one shared database.
This page is the honest head-to-head. We'll walk through when to pick Sage Intacct, when to pick ERPClaw, and where each falls short of the other. If you want the broader category map, start with our AI-native ERP framework.

## TL;DR comparison
The short version. Eight rows, side by side. The full reasoning is in the sections below.
| Dimension | Sage Intacct | ERPClaw
| Architecture | 1990s general ledger product with Sage Copilot added as a paid module | AI-native from line one; action layer is the primary surface
| Delivery | Sage cloud only (SaaS, multi-tenant) | Self-host (Docker, SQLite or PostgreSQL); your hardware, your data
| Pricing | Core from around $15,000 per year, plus add-on modules (fixed assets, subscription billing, multi-entity), plus implementation | $0 forever, unlimited seats, every module included
| ERP scope | Mid-market finance plus add-on modules; deeper at finance, lighter at full ERP | Full ERP (accounting, inventory, manufacturing, payroll, Stripe, Shopify) plus 14 industry verticals
| Multi-currency | Yes (mature multi-entity FX, ASC 830, ASC 842 lease) | 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED); invoice currency equals payment currency, no FX guessing
| Audit trail | Yes; AICPA preferred-provider with audit-firm familiarity | SHA-256 chain hash plus 12-step GL invariant validation pipeline
| Open source | Proprietary, closed source | Free &amp; Open Source, full source on GitHub
| Best for | Mid-market finance teams ($50M to $500M revenue) who need vendor-managed SaaS and audit-firm familiarity | Solo founders to mid-market who want AI-native scope without per-module pricing

If your audit committee requires AICPA-endorsed vendor SaaS and your finance team runs 5+ subsidiaries with monthly consolidation, Sage Intacct is the safer choice. If you want AI-native scope across every business function with no per-module tax, open source, and your data on your hardware, that's ERPClaw.

## When Sage Intacct is the right choice
Credit where due: Sage Intacct is a real mid-market finance platform with AICPA preferred-provider status and deep multi-entity tooling. Here are three situations where it beats us today.

### You need deep multi-entity consolidation maturity.
Sage Intacct's multi-entity consolidation, ASC 842 lease accounting, project accounting, and dimensions model are battle-tested for mid-market finance teams with 5 to 50 subsidiaries. ERPClaw covers the primitives, but Intacct's UX for that specific profile (close cockpit, intercompany eliminations, multi-currency consolidation) is more refined today. If your CFO measures success in days-to-close across many entities, Intacct's polish on that workflow matters.

### Your auditor or audit committee prefers AICPA-endorsed software.
Sage Intacct is the AICPA's preferred-provider for accounting software. Mid-market audit firms are familiar with it, and that familiarity reduces audit friction. If your audit committee asks "is this an AICPA-approved system?", Sage Intacct's commercial structure answers that question directly. ERPClaw is Free &amp; Open Source with documented controls and a chain-hashed audit trail (the same defensibility model auditors accept for Linux on production servers), but it's not AICPA-endorsed as a vendor.

### You want a vendor-managed SaaS with a contractual SLA.
Sage Intacct is hosted by Sage with account managers, premium support tiers, and a contractual SLA. If your finance team has zero ops capacity and your CFO mandates a vendor-managed counterparty, that delivery model has real value. ERPClaw is open source plus the co-founder team; for some governance frameworks, that mismatches the requirement.

## When ERPClaw is the right choice
Five situations where the structural choice points to us. Source on github.com/avansaber/erpclaw if you want to read before you decide.

### You want AI-native, not a Copilot add-on.
Sage Copilot is a chatbot added to a 1990s general ledger product as a paid module. The architecture predates the AI era. ERPClaw was designed AI-native from line one: every accounting action is invokable from a natural-language prompt with invariants enforced at write time. That structural choice cannot be retrofitted onto a 1990s codebase.

### You want $0 forever, not $15K+ per year plus add-on modules.
Sage Intacct's published core is roughly $15,000 per year, but the add-on modules (fixed assets, subscription billing, multi-entity consolidation) are commonly priced higher than the core, plus implementation. ERPClaw is $0 forever, every module included: immutable GL, ASC 606 revenue recognition for SaaS, multi-entity, inventory, manufacturing, payroll, Stripe, Shopify, plus 14 industry verticals. One install, one license, no upsell.

### You want full ERP, not finance-only-plus-add-ons.
Sage Intacct's scope is mid-market finance with add-on modules layered on. ERPClaw covers AR, AP, GL, payroll (US: W-2, 1099, NACHA, FICA, FUTA, SUTA), inventory, tax, Stripe (live on Stripe Marketplace), Shopify (OAuth Token Exchange) in one shared database. If you need to invoice, ship, pay employees, and reconcile in one place, that scope difference is the conversation.

### You want to own your data and your code.
Sage Intacct stores your books on Sage's cloud. ERPClaw runs on your hardware (laptop, server, anywhere) in SQLite or PostgreSQL with documented schemas. Migration off Sage Intacct requires CSV exports and re-mapping; with ERPClaw, the export pattern is `cp data.sqlite somewhere-else.sqlite`. Your data is portable from day one because the source is yours.

### You want any business action invokable from natural language.
ERPClaw's AI-native architecture means the action layer is the API. Type "add Acme as a customer for $5K MRR starting June 1" and the AI invokes add-customer, add-subscription, the deferred-revenue GL postings, and the ASC 606 schedule in one transaction with a full audit row. Sage Copilot is scoped to specific finance lookups; ERPClaw's chat is the primary interface for the whole business.

For the deeper open-source argument, see open-source AI accounting. For the AI-native framework, see AI-native ERP.

## Pricing
Concrete cost items over a real horizon. Sage Intacct's exact quote depends on tier, modules selected, user count, and entity count; the figures below are widely reported public ranges, not a binding quote.
| Cost item | Sage Intacct | ERPClaw
| Core license (per year) | From around $15,000, plus per-seat scaling | $0 forever, unlimited seats
| Add-on modules | Fixed assets, subscription billing, multi-entity consolidation, project accounting commonly priced higher than the core | Every module included always
| Implementation | $10K to $100K+, 3 to 12 months | 5-minute install, no implementation engagement
| ASC 606 / ASC 842 | Included or add-on depending on tier | ASC 606 built in via Stripe; ASC 842 lease accounting built in
| Source code access | Closed | Free &amp; Open Source: fork it, audit it, run it
| Where it runs | Sage cloud (your data on Sage's servers) | Your hardware (laptop, server, anywhere)

For a $10M to $100M revenue mid-market finance team, the multi-year cost differential between Sage Intacct and ERPClaw is significant: Intacct typically runs to six figures across core license, add-on modules, and implementation over three years; ERPClaw is zero. See our pricing page for the full breakdown.

## Migration: switching to ERPClaw
Switching ERPs is friction. Here is the honest path.
From Sage Intacct. Export your chart of accounts, customers, vendors, items, and transaction history via Intacct's CSV exports. Import via ERPClaw's `import-chart-of-accounts` and `import-opening-balances` actions. Intacct supports CSV export across all main objects, so the path is mechanical. The dimensions model maps cleanly to ERPClaw's tagging model.
From QuickBooks Online or Xero. Same CSV path. ERPClaw's import actions accept the standard chart-of-accounts and trial-balance shapes. There is no one-click migrator yet; we are honest about that.
Stripe history. ERPClaw's Stripe integration backfills directly from the Stripe API with ASC 606 schedules generated as the data lands. Connect, run the backfill action, the GL entries post with full audit rows.
Shopify history. Same model. The Shopify integration (OAuth Token Exchange via App Bridge) pulls orders, payouts, and adjustments straight from Shopify.
Honest gap. A multi-entity Intacct tenant with 10+ subsidiaries on multi-book with ASC 842 lease accounting is not a weekend port. Plan for a phased migration: one entity at a time, parallel-run for a quarter, then cut over. Reach us through support if you want a sized estimate, and see the product working at /demo/.

## What Sage Intacct does better than ERPClaw today
Sage Intacct wins on these dimensions today, and pretending otherwise would be insulting to a serious peer.

- Mid-market consolidation maturity. Multi-entity, multi-book, ASC 830 FX gain/loss at scale, and the close cockpit UX are more polished than ERPClaw's today for finance teams running 5 to 50 subsidiaries.
- AICPA preferred-provider plus audit-firm familiarity. Mid-market audit firms are trained on Sage Intacct; that reduces audit friction at the margin.
- Native dimensions model. Sage Intacct's dimensions architecture (location, department, project, customer as orthogonal slices) is widely cited as best-in-class for mid-market financial reporting.
- Polished web UI. ERPClaw is CLI plus chat first; the webclaw web dashboard exists but is limited for Stripe and Shopify today.

## What ERPClaw does that Sage Intacct doesn't
The structural wins. These are not feature checkmarks; they are architectural choices Sage Intacct cannot retrofit without rebuilding on top of a 1990s general ledger product.

- AI-native architecture. The action layer is the AI's API; every action invokable from prompt. Sage Copilot is a chatbot module added to a 1990s general ledger product.
- Free &amp; Open Source. Fork, contribute, audit. Sage Intacct is proprietary, closed source.
- Self-host. Your data on your hardware. Sage Intacct is Sage cloud only.
- $0 forever, every module included. No add-on tax on fixed assets, subscription billing, or multi-entity consolidation.
- 5-minute install. Sage Intacct implementations run 3 to 12 months.
- Full ERP scope plus 14 industry verticals. Sage Intacct's scope is mid-market finance plus add-ons.
- Multi-currency in 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED). Invoice currency equals payment currency, no FX guessing.
- 12-step GL invariant validation on every posting plus SHA-256 chain-hashed audit trail. Bad entries cannot reach the books.
- Stripe integration live on Stripe Marketplace. Shopify integration shipped at v1.1.3 (OAuth Token Exchange via App Bridge).
- Patent pending plus trademark filed. Long-term durability signal.

## Frequently asked questions

### How does Sage Intacct add-on pricing actually work?
Sage Intacct's published-price core is roughly $15,000 per year plus implementation, but the add-on modules (fixed assets, subscription billing, multi-entity consolidation, project accounting) are commonly priced higher than the core. The math grows fast for a mid-market finance team. ERPClaw includes immutable GL, ASC 606 revenue recognition for SaaS, and multi-entity in the open-source build, with no add-on tax.

### Is ERPClaw really comparable to Sage Intacct on accounting depth?
ERPClaw implements the primitives mid-market finance leaders care about: immutable double-entry GL, multi-entity consolidation with eliminations, ASC 606 revenue recognition for SaaS, ASC 842 lease accounting, an audit-ready chain-hashed trail, and a posting validation pipeline that checks every entry. The accounting layer is solid. Where Sage Intacct still leads is transactional FX revaluation at scale (intentionally out of ERPClaw's scope) and automatic consolidation-time translation, which ERPClaw records as a deliberate entry rather than computing, plus certain industry-specific dashboards. If those are blockers for your team, evaluate side by side.

### What's the AI difference between Sage Copilot and ERPClaw?
Sage Copilot is a chatbot added to a 1990s general ledger product as a paid module. The architecture predates the AI era. ERPClaw was designed AI-native from line one: every accounting action is invokable from a prompt, with constitutional invariants enforced at write time. The 5-trait test on our AI-native ERP pillar makes the difference concrete.

### Can my CFO defend choosing open-source for finance?
Yes. ERPClaw is Free &amp; Open Source with a chain-hashed audit trail, extensive automated test coverage, and source you can read line by line. SOC 2-style controls are documented. Open source is increasingly preferred by auditors over SaaS black boxes because the controls are observable. Your CFO defends it the same way they defend Linux on production servers. If your audit committee specifically requires an AICPA-endorsed vendor, Sage Intacct's commercial structure fits that mandate directly; ERPClaw does not.

### How does ERPClaw's mid-market positioning differ from Sage Intacct?
Sage Intacct targets the $50M to $500M revenue mid-market with managed SaaS and a six-figure entry cost (license plus add-on modules plus implementation). ERPClaw targets solo founders to mid-market at $0, self-hosted, with the same finance primitives plus inventory, manufacturing, HR, payroll, and 14 industry verticals. The structural moats matter more at the mid-market than at enterprise.

### What does ERPClaw not do that Sage Intacct does?
ERPClaw's web dashboard is included free and currently in beta; most users work through the chat interface, which stays the fastest path today. Transactional FX revaluation at scale is intentionally out of scope, not a roadmap commitment. Automatic consolidation-time translation is not shipped either: translating a subsidiary is a deliberate entry today. 24/7 vendor-managed support with a contractual SLA is not available; support is community-tier through GitHub plus the co-founder team. AICPA preferred-provider status is not something an open-source project applies for.

### Can ERPClaw replace Sage Intacct entirely?
For most mid-market use cases, yes. Both cover AR, AP, GL, multi-entity, multi-currency, ASC 606, and ASC 842 lease accounting. ERPClaw also covers inventory, payroll, tax, Stripe, and Shopify natively. The honest gaps: transactional FX revaluation at scale (intentionally out of scope) and AICPA-endorsed vendor status. If those specifically block your buyer, Sage Intacct is the cleaner fit today. For everything else (scope, source code access, self-host, $0), ERPClaw is the cleaner answer.

## Where to go next
Free forever. Install in 5 minutes. Every module included.
Talk to a co-founder Read the AI-native framework
Or browse the source on github.com/avansaber/erpclaw.

---

## SAP Business One Alternative: Free AI-Native ERP
URL: https://www.erpclaw.ai/compare/sap-business-one/

# ERPClaw vs SAP Business One
SAP Business One is per-user licensed plus partner implementation. ERPClaw is open-source AI-native at $0 forever.
Same primitives. AI-native by design. Free.

| Feature | SAP Business One | ERPClaw
| Pricing | $94-$148+/user/mo plus $1.5K-$3K/user one-time license | $0 forever
| Implementation | $30K-$200K via SAP partner | $0 (5-min install)
| Implementation time | 3-9 months | 5 minutes
| Modules | Per-tier (Professional, Limited) | All 46 included always
| Multi-entity | Yes (Professional) | Yes
| AI capabilities | Partner add-ons via SDK (SAP Joule is in S/4HANA, not Business One) | AI-native (built into the action layer)
| Customization | SDK, partner-driven | Spec-first; any LLM regenerates code
| Self-hosted | Yes (on-prem option) or cloud | Yes
| Database | SAP HANA or Microsoft SQL Server | Database-agnostic via PyPika; SQLite default, PostgreSQL fully supported
| Open source | Proprietary | Free &amp; Open Source
| Vendor lock-in | Significant (SAP ecosystem + partner network) | None (copy the database file and walk away)

SAP Business One + Joule facts verified 2026-05 from SAP News Q1 2026 release highlights (news.sap.com) and SAVIC Technologies industry coverage.

## Key differences

### $0 vs SAP partner economics
SAP Business One has a license fee plus per-user subscription plus a partner implementation. Combined annual cost for a 10-person team commonly lands $30K to $80K. ERPClaw is $0 forever, every module, with a 5-minute install.

### AI-native architecture from line one
SAP Joule is the assistant for S/4HANA, SuccessFactors, and Ariba; it is not embedded in Business One. Q1 2026 SAP announcements (Joule Studio GA, 40+ specialized Joule agents, the SAP-ABAP-1 and SAP-RPT-1 foundation models) all target S/4HANA and BTP, not B1. AI in Business One today is partner-built add-ons via the SAP B1 SDK on top of an ERP architecture that predates the AI agent era. ERPClaw was rebuilt around AI; the action layer is AI-native from line one. The 5-trait test on /ai-native-erp/ scores both layer by layer.

### Spec-first regen vs SAP SDK
Business One customizations live in SAP's SDK and require partner engagement. ERPClaw customizations live in SKILL.md specs that any LLM regenerates from. No partner required.

### Open source vs SAP ecosystem
Business One is closed source even when self-hosted. ERPClaw is Free &amp; Open Source with full source on GitHub; you can read, modify, fork, or audit the code.

## Frequently asked questions

### How does SAP Business One pricing actually work?
Business One uses a hybrid pricing model: a one-time license fee per user (around $1,500 to $3,000 per Professional named user, lower for Limited users) plus a per-user maintenance subscription ($94 to $148 per user per month). Add the SAP partner implementation cost ($30K to $200K typical). ERPClaw is $0 forever with no license fee, no maintenance subscription, and no partner requirement.

### Is SAP Business One AI-native?
No. SAP's flagship AI assistant Joule is built for S/4HANA, SuccessFactors, and Ariba; it is not embedded in Business One. AI in Business One is partner-built add-ons via the SAP B1 SDK on top of an ERP architecture that predates the AI agent era. ERPClaw is AI-native: the agent posts journal entries autonomously across every module. The 5-trait test on /ai-native-erp/ scores them layer by layer.

### Does SAP Business One include SAP Joule in 2026?
No. SAP Joule is embedded in SAP S/4HANA Cloud, SuccessFactors, and Ariba. SAP Business One's AI surface in 2026 is partner-built add-ons via the SAP B1 SDK. The 40+ Joule agents and 2,500+ Joule Skills announced at Q1 2026 are for S/4HANA, not Business One. ERPClaw includes AI-native architecture in every install with no separate add-on or partner required.

### Can ERPClaw replace SAP Business One for SMB manufacturing?
Most SMB manufacturing setups: yes. ERPClaw covers items, BOMs, work orders, FIFO and weighted-average costing, multi-warehouse inventory with reservations, and US-GAAP-compliant immutable GL. Where Business One still leads is some industry-specific add-ons in the SAP partner ecosystem and certain compliance modules outside US scope.

### What about migrating from SAP Business One to ERPClaw?
Business One exports cleanly to CSV per Business Object. The ERPClaw AI agent reads the export and runs the import actions. Custom UDFs and SDK customizations need to be re-expressed as SKILL.md specs; the AI helps that translation. Plan 4-8 weeks elapsed including parallel run; the SAP partner is no longer in the loop.

### Is SAP support worth the partner cost?
For some businesses, yes. SAP partners offer industry-specific expertise and 24/7 support. For most SMB Business One deployments, the partner cost is a tax on the buyer rather than a benefit. ERPClaw support is community-tier (GitHub, Discord) plus optional paid services from AvanSaber Inc. The math at SMB scale strongly favors ERPClaw.

---

## SAP Business One vs ERPClaw on AI-Native Architecture
URL: https://www.erpclaw.ai/compare/sap-business-one/ai-native-architecture/

# SAP Business One vs ERPClaw: AI-Native Architecture
Is SAP Business One AI-native in 2026? The honest answer matters when you're picking an ERP.

## The question, plainly
As of Q1 2026, SAP Business One's AI surface is partner add-ons via the SAP B1 SDK. Joule, SAP's headline AI assistant, lives in S/4HANA Cloud, SuccessFactors, and Ariba. It is not embedded in Business One.
That distinction matters. Joule's Q1 2026 expansion (Joule Studio GA, 40+ specialized agents, the SAP-ABAP-1 and SAP-RPT-1 foundation models) targets the S/4HANA and BTP customer base. Business One customers are downstream of that roadmap.
Sources: news.sap.com Q1 2026 release highlights, SAVIC Technologies industry coverage.

## The 5-trait AI-native test, applied to SAP Business One
Five questions separate AI-native ERP from AI-decorated ERP. Here is how Business One scores in 2026.

### AI writes to GL directly
No
Partner add-ons surface suggestions; a human always approves the posting. The GL write path is human-driven by design.

### Schema designed for AI agents
No
The SAP Business One schema predates the AI agent era. SDK extensions added later sit on top of a model built for forms and screens, not autonomous action.

### Native action layer
Partial
The SAP B1 SDK is a programmatic surface, but it is shaped for human-driven workflows and partner customizations, not AI-first call patterns.

### Pre-write invariant enforcement
Partial
SAP enforces GL invariants, but checks land at batch boundaries rather than per-write. An AI agent posting in tight loops cannot rely on per-call validation.

### Single AI tier (no gating)
No
AI capability in Business One requires a partner contract. There is no shipped, in-box AI tier for B1 customers in 2026.

## The same 5-trait test, applied to ERPClaw
Same questions, different architecture.

### AI writes to GL directly
Yes
The AI agent posts journal entries through the same action layer a human uses. No staging, no approval gate by default.

### Schema designed for AI agents
Yes
Every table, every action, every invariant was designed assuming an LLM is the primary caller. SKILL.md files are the contract.

### Native action layer
Yes
Every module expose typed actions. The action surface is the same one the AI calls, the CLI calls, and the API calls.

### Pre-write invariant enforcement
Yes
12-step pre-write validation runs on every GL posting before the transaction commits. The AI cannot post an invalid entry.

### Single AI tier
Yes
$0 forever. All AI capabilities included. No partner contract, no subscription, no per-user fee.

## Capability-by-capability on AI-native dimensions
| Dimension | SAP Business One | ERPClaw
| AI agent posts to GL | No (human approves) | Yes
| Native action layer for AI | SDK (human-shaped) | Yes (AI-first)
| Pre-write invariants | Batch-level | 12-step per-write
| Schema designed for agents | No | Yes
| AI tier | Partner add-on | Included ($0)

## What this means for an SMB founder
If you want AI-native at the SAP Business One price point in 2026, you have real options. Each one has tradeoffs worth understanding before you sign.
Odoo Enterprise ships AI features (lead scoring, OCR, AI Fields, semantic search, agentic roadmap), all gated behind a per-user subscription. ERPNext is open source, capable, and active, with AI capability coming from third-party Frappe apps like NextAI, ChatNext, and ChangAI rather than a native action layer. ERPClaw is AI-native by design, Free & Open Source, and $0 forever.
The right pick depends on whether you want AI included or tiered, whether you can run an open-source stack, and how much you value an action layer the AI was designed for.

## Frequently asked questions

### Is SAP Business One AI-native in 2026?
No. As of Q1 2026, SAP's headline AI product Joule is embedded in S/4HANA Cloud, SuccessFactors, and Ariba. SAP Business One's AI surface is partner add-ons built on the B1 SDK. That is AI decoration on top of a pre-AI architecture, not AI-native.

### When will SAP Joule come to Business One?
SAP has not announced a Joule roadmap for Business One in 2026 public communications. Q1 2026 SAP News coverage and SAVIC Technologies industry reporting show Joule expansion targeting S/4HANA Cloud, SuccessFactors, and Ariba. B1 customers should treat any Joule arrival as undated.

### Can I get AI on SAP B1 today?
Yes, via SAP partner add-ons through the SAP B1 SDK. These typically wrap third-party LLM APIs around B1 data and require a partner engagement to deploy and maintain. The economics differ from an in-box AI tier.

### What's the alternative if I want true AI-native at the SMB tier?
At the B1 price point in 2026, you have a few options. Odoo Enterprise gates AI behind a per-user subscription. ERPNext relies on third-party Frappe apps for AI. ERPClaw is AI-native, $0 forever, and self-hosted.

## Read the architecture, then decide
The 5-trait test is the framework. Apply it to any ERP you're evaluating.
Try the demoRead the 5-trait test

Related: the parent SAP Business One comparison, the AI-native ERP framework, or the 5-trait research note.

---

## Workday alternative: ERPClaw is open source, AI-native, free
URL: https://www.erpclaw.ai/compare/workday/

# ERPClaw vs Workday
Workday is enterprise HCM and Financials. ERPClaw is the open-source AI-native option for SMB-to-mid-market. Honest about where each fits.
For the right tier. AI-native by design. Free.

| Feature | Workday | ERPClaw
| Pricing | Contact sales (founder-reported $100-$200/employee/year for HCM) | $0 forever
| Implementation | $300K-$5M+ via Workday partner | $0 (5-min install)
| Implementation time | 9-24 months | 5 minutes
| Modules | HCM, Financial Management, Adaptive Planning (separate) | All 46 included always
| Multi-entity | Yes (Financial Management) | Yes
| AI capabilities | Workday AI / Workday Illuminate (chat sidebar on enterprise HCM) | AI-native architecture
| Customization | Workday Studio, partner-driven | Spec-first; any LLM regenerates code
| Self-hosted | No (Workday cloud only) | Yes (SQLite or PostgreSQL)
| Open source | Proprietary | Free &amp; Open Source
| Vendor lock-in | High (cloud-only, custom data model) | None

## Key differences

### $0 vs $100-$200 per employee per year
Workday HCM pricing is contact-sales but founder-reported quotes commonly land $100-$200 per employee per year, with HCM plus Financial Management plus Adaptive Planning combined easily clearing $200K per year for a 1,000-person company. ERPClaw is $0 forever, all modules included.

### Cloud-only lock-in vs self-hosted portability
Workday is cloud-only; your data lives on Workday's servers. Migration off Workday is a notoriously expensive engagement. ERPClaw is self-hosted on your hardware; the export pattern is literally cp data.sqlite somewhere-else.sqlite.

### AI-native instead of Workday Illuminate
Workday's AI offerings (Illuminate, Workday AI) are chat sidebars on top of the existing enterprise HCM and Financials products. ERPClaw was rebuilt around AI. The 5-trait test on /ai-native-erp/ scores both layer by layer.

### We don't replace Workday at Fortune 500
Workday is built for $1B+ revenue enterprises with multi-thousand-employee deployments. ERPClaw is built for solo founders to mid-market. We are honest: at the Fortune 500 scale, Workday's depth in HCM, multi-entity consolidation, and global compliance still matters. Where ERPClaw replaces Workday is at the SMB-to-mid-market entry tier where Workday pricing becomes brutal.

## Frequently asked questions

### Is ERPClaw really comparable to Workday?
Honest answer: not at the Fortune 500 scale Workday is built for. Workday's HCM depth, multi-entity consolidation, and global compliance modules are real engineering investments. ERPClaw covers the SMB-to-mid-market entry tier (under $300M revenue, under 1,000 employees) where Workday pricing becomes prohibitive. At that tier, ERPClaw covers the same primitives at $0 with structurally newer architecture.

### What does Workday actually cost?
Workday pricing is contact-sales and varies wildly by employee count, modules, and contract terms. Founder-reported numbers commonly land $100 to $200 per employee per year for HCM alone; the full HCM plus Financial Management plus Adaptive Planning stack at a 1,000-person company commonly clears $200K per year plus a $300K to $5M+ implementation. ERPClaw is $0 forever.

### Can ERPClaw handle enterprise payroll the way Workday does?
ERPClaw includes a US payroll module (W-2, 1099, NACHA, FICA, FUTA, SUTA). What it does not have today is multi-country payroll for 50-state-plus-international workforces or the deep HCM workflow modeling that Workday is built around (talent management, performance reviews, succession planning at enterprise scale). For SMB-to-mid-market US-focused payroll, ERPClaw is sufficient. For Fortune 500 global HCM, Workday is still the answer.

### Why would a Workday customer migrate to ERPClaw?
Two scenarios. First, a Workday customer has been acquired or restructured down to a smaller entity that no longer justifies Workday's enterprise pricing; ERPClaw is the right-sized replacement. Second, a Workday customer is fed up with the implementation timeline (9-24 months for new features) and wants AI-native architecture they can self-host. Both are real but specific situations.

### Is Workday AI-native?
No. Workday Illuminate and Workday AI are chat-and-suggestion sidebars on top of the existing HCM and Financials data model. The architecture predates the AI agent era. ERPClaw is AI-native by design from line one. The 5-trait test on /ai-native-erp/ scores them layer by layer.

---

## Xero Alternative - Free Open Source ERP With AI Chat
URL: https://www.erpclaw.ai/compare/xero/

# ERPClaw vs Xero
Xero charges $13-78/mo. ERPClaw is free.
Full ERP scope. Open source. Chat interface.

| Feature | Xero | ERPClaw
| Price | $13-78/month | $0 forever
| Users included | 1-unlimited (tier-dependent) | Unlimited (always)
| Multi-entity | Per-organization subscription | Unlimited entities free
| Inventory | Add-on or premium tier | Built-in (multi-warehouse)
| Manufacturing | No | Built-in (BOM, MRP, work orders)
| HR / Payroll | Add-on (Xero Payroll) | Built-in (US payroll, FICA, W-2)
| CRM | No | Built-in
| AI capabilities | Just Ask Xero (chat sidebar; human approves every action) | AI-native (action layer is the primary surface)
| Database | Closed proprietary cloud storage | Database-agnostic via PyPika; SQLite default, PostgreSQL fully supported
| Self-hosted | No (cloud only) | Yes (your server)
| Open source | Proprietary | Open source (Free &amp; Open Source)
| Stripe + Shopify | Add-on apps required | Built-in deep integrations
| Multi-currency | Yes (premium tier) | USD in v1

Xero + Just Ask Xero facts verified 2026-05 from Xero's public product pages and 2025-2026 product announcements.

## Key differences

### Beyond accounting
Xero is excellent accounting software. ERPClaw is a full ERP. Inventory, manufacturing, HR, CRM - all included.

### No per-user pricing
Xero charges per organization with per-tier user limits. ERPClaw is unlimited users always.

### Self-hosted means audit-ready
Your books live on your server. Backups are file copies. Auditors can query the SQLite database directly.

### AI-native, not AI-bolted-on
Just Ask Xero is a chat assistant in front of the same Xero forms; a human still posts. ERPClaw was designed around chat from day one, and the AI is the primary user of the action layer.

---

## Open-Source Python ERP for Developers
URL: https://www.erpclaw.ai/developers/

# Build on OpenClaw
ERPClaw is a modular skill suite on the OpenClaw platform. Every module is an independent skill with a Python script, a SKILL.md manifest, and full test coverage.

## Architecture
Six layers, top to bottom. Every request flows through the same path.
Layer 1

### Telegram / WhatsApp / Discord / Web UI

Users interact through any messaging platform or the Webclaw browser dashboard. Natural language in, structured results out.

Layer 2

### OpenClaw (AI Router)

The AI runtime reads SKILL.md manifests, maps user intent to skill actions, and formats JSON responses into human-readable output.

Layer 3

### SKILL.md (Manifest)

A declarative YAML file that describes the skill's name, version, actions, parameters, and progressive disclosure tiers.

Layer 4

### ERPClaw OS

The self-improving engine. Constitutional framework, module generation, semantic correctness, and gap detection. Governs all changes to the system.

Layer 5

### db_query.py --action {name}

A single Python script per skill. Every action is routed through the --action flag. Output is always JSON to stdout. PyPika query builder for DB portability.

Layer 6

### SQLite or PostgreSQL

All modules share one database. SQLite (default) or PostgreSQL (enterprise). WAL mode, foreign keys enforced, PyPika abstraction layer for DB-agnostic queries.

## Skill API Quick Reference
Every skill follows the same two-file pattern: a SKILL.md manifest and a db_query.py script.

### SKILL.md Format

`name: erpclaw-selling
version: "1.0.0"
description: Sales orders, invoices, delivery notes
author: AvanSaber Inc.
scripts:
- name: db_query.py
description: Sales management
arguments:
- name: action
description: Action to execute
required: true
actions:
- name: add-customer
description: Create a new customer
tier: 1
- name: list-customers
description: List all customers
tier: 1`

### Action Interface

`# Every action follows the same pattern:
python db_query.py --action add-customer \
--customer_name "Acme Corp" \
--customer_group "Commercial" \
--credit_limit 50000

# Output is always JSON:
{
"customer_id": "abc-123-def",
"customer_name": "Acme Corp",
"naming_series": "CUST-2026-00001",
"status": "active"
}`

Modules

3,234
Actions

v4.15.0
Current release

Signed
Module registry

## Database Schema
One schema in a single database. Every module has its own table namespace. SQLite (default) or PostgreSQL (enterprise).
| Module | Description
| Core (setup, GL) | Companies, users, accounts, GL entries, periods
| Financial (journals, payments, tax, reports) | Journal entries, payment allocation, tax rules
| Supply Chain (inventory, selling, buying) | Items, warehouses, stock ledger, SO/PO/invoices
| Manufacturing | BOMs, work orders, job cards, MRP
| HR & Payroll | Employees, leave, attendance, salary, expenses
| CRM & Support | Leads, opportunities, campaigns, issues, SLAs
| Projects & Quality | Projects, tasks, timesheets, inspections
| Assets & Billing | Fixed assets, depreciation, subscriptions, metering
| AI & Analytics | KPIs, anomalies, forecasts, scores
| All of it, one schema | A single database with WAL mode and indexes to match (SQLite or PostgreSQL)

## Quick Start
From zero to a working skill in three steps.

### Clone & Install
Clone the setup skill, initialize the database, and load demo data.
`git clone https://github.com/avansaber/erpclaw-setup
cd erpclaw-setup
python scripts/db_query.py --action init-db
python scripts/db_query.py --action seed-demo-data`

### Test
Run the test suite to verify everything works.
`cd tests && pytest -v`

### Develop
Create your own skill following the standard two-file pattern.
`# Create a new skill
mkdir my-skill && cd my-skill
# Add SKILL.md + scripts/db_query.py
# Follow the action pattern: --action my-action --param value`

---

## Documentation: Install, Stripe, Shopify Setup
URL: https://www.erpclaw.ai/docs/

# ERPClaw Documentation
Everything you need to install ERPClaw, set up your books, and connect Stripe and Shopify. New to ERPClaw? Start with install in 5 minutes.

## Core ERP
15 pages
Install ERPClaw, set up your first company, and learn how the double-entry GL works.

- Install ERPClaw in 5 minutes
Step-by-step install guide for the ERPClaw core. Covers OpenClaw + ClawHub install, database init, first commands, and verification.

- Accounting basics in ERPClaw
How the chart of accounts, journal entries, and GL posting work in ERPClaw. Double-entry, immutable submitted entries, the 12-step validation pipeline.

- Purchasing in ERPClaw
How purchase orders, goods receipts, vendor invoices, and payments flow in ERPClaw. Three-way matching, GL postings at each step, supplier setup.

- Import a bank statement
Upload an OFX, CAMT.053, MT940, or BAI2 statement, auto-match the lines to open invoices and payments, review what is left, and re-import safely without doubling up.

- Reports by department, project, or any dimension
Tag entries with a department, project, or any dimension you define, then ask for a profit and loss split by it, with untagged amounts shown separately.

- Reserve stock for an order
Hold a quantity for one order so another order cannot quietly consume it. How reservations are checked, what gets refused, and how to release stock you no longer need to hold.

- Putaway rules and pick lists
Route incoming goods to the right warehouse automatically, and fill orders from a guided pick list that reserves stock as it goes and becomes a delivery note when complete.

- Item substitutes and alternatives
Set backup items once at the item level so the system already knows the ranked alternative when your first choice runs short, including for manufacturing bills of materials.

- Install ERPClaw on the Hermes Agent runtime
Experimental Hermes support via a GitHub tap: the two-command install, why the full skill identifier is mandatory, and the manual clone alternative.

- Subcontract production to an outside manufacturer
Send raw materials out, receive finished goods back, and let the books roll material and subcontracting costs into the finished item automatically.

- Contacts and companies
Keep people and the businesses they work for as real records, link them, merge duplicates safely, and turn a contact into a paying customer in one step.

- Sales tasks and follow-ups
Track follow-ups with due dates and priorities, linked to the deal, contact, or customer they belong to, so nothing promised in a call dies in a notebook.

- Sales pipelines with your own stages
Build deal stages that match how you actually sell, not a fixed list, and see every open opportunity grouped by where it really stands.

- Saved views and search across Sales
Save a filtered list you actually reuse, share it with the team, and search across every sales record at once instead of remembering where something lives.

- Bring your sales data in, and take it out, as CSV
Import leads, contacts, companies, and deals from a spreadsheet with explicit duplicate handling, and export any list back out. Your data stays yours.

## Stripe
6 pages
Sync Stripe payments to a real general ledger. ASC 606 revenue recognition and Connect platform fees included.

- Integration Install Walkthrough
Step-by-step Stripe accounting integration install for ERPClaw. From fresh Stripe account to first synced charge in under 5 minutes.

- Integration Architecture in ERPClaw
Technical architecture of the ERPClaw Stripe connector. Direct API access from your machine. No middleware. Optional webhook receiver.

- Integration Troubleshooting
Fix common ERPClaw Stripe integration issues. API key errors, sync failures, webhook 401s, payout reconciliation gaps, ASC 606 imbalances, MRR misreports.

- ASC 606 Revenue Recognition for SaaS
How ERPClaw handles Stripe ASC 606 revenue recognition. Deferred revenue, monthly recognition, mid-cycle changes, refunds, and the journal entries that prove it works.

- ERPClaw Connect Fees
Configure ERPClaw for Stripe Connect direct charges, destination charges, and separate charges and transfers, including fields, events, refunds, transfers, and payouts.

- MRR Reports
How ERPClaw computes MRR, ARR, churn, expansion, and contraction from your Stripe subscription data. Cohort retention, net new MRR, and the math that powers SaaS reports.

## Shopify
6 pages
Pair ERPClaw to your Shopify store. Order sync, GDPR webhooks, and OAuth pairing walkthrough.

- Integration Install Walkthrough
Step-by-step Shopify accounting integration install for ERPClaw. From fresh Shopify store to first synced order in under 10 minutes.

- Integration Architecture in ERPClaw
Technical architecture of the ERPClaw Shopify connector: Cloudflare Pages plus Worker plus your ERPClaw. Merchant data never transits our servers.

- Integration Troubleshooting
Fix common ERPClaw Shopify integration issues. Pairing code expired, OAuth scope errors, sync failures, status card stuck, webhook 401s, and reinstall flow.

- OAuth Pairing
The ERPClaw Shopify OAuth pairing flow: six-character codes, single-use redemption, per-shop HMAC, and App Bridge session tokens. Tokens never persist.

- GDPR Webhooks
How the ERPClaw Shopify integration handles the four mandatory GDPR webhooks: data request, customer redact, shop redact, and app uninstalled.

- App Store vs Custom App
Comparison of the two ERPClaw Shopify install paths: Shopify App Store OAuth (recommended) vs Custom App with shpat_ token (power-user).

Learn

## Continue learning about ERPClaw
Move from installation and operations into architecture, research, product guidance, quality, security, and support.
DocumentationInstall and operate ERPClawDevelopersArchitecture, APIs, and extension pointsBlogProduct, accounting, and ERP guidanceResearchOpen methods and original studiesAI ERP guideUnderstand the categoryAI-native ERP guideEvaluate architecture and controlsOpen source AI accountingA practical market guideQualityHow ERPClaw is testedSecurityDeployment and data controlsSupportGet help from AvanSaber

---

## Core Docs
URL: https://www.erpclaw.ai/docs/core/

# Core docs
Install, accounting model, purchasing flow, and the rules behind the 12-step GL validation pipeline. Start here.
Install in 5 minutesPricing

## All core docs

-

### Install ERPClaw in 5 minutes
install
Step-by-step install guide for the ERPClaw core. Covers OpenClaw + ClawHub install, database init, first commands, and verification.

-

### Accounting basics in ERPClaw
accounting
How the chart of accounts, journal entries, and GL posting work in ERPClaw. Double-entry, immutable submitted entries, the 12-step validation pipeline.

-

### Purchasing in ERPClaw
purchasing
How purchase orders, goods receipts, vendor invoices, and payments flow in ERPClaw. Three-way matching, GL postings at each step, supplier setup.

-

### Import a bank statement
import-bank-statement
Upload an OFX, CAMT.053, MT940, or BAI2 statement, auto-match the lines to open invoices and payments, review what is left, and re-import safely without doubling up.

-

### Reports by department, project, or any dimension
reports-by-dimension
Tag entries with a department, project, or any dimension you define, then ask for a profit and loss split by it, with untagged amounts shown separately.

-

### Reserve stock for an order
stock-reservations
Hold a quantity for one order so another order cannot quietly consume it. How reservations are checked, what gets refused, and how to release stock you no longer need to hold.

-

### Putaway rules and pick lists
putaway-and-pick-lists
Route incoming goods to the right warehouse automatically, and fill orders from a guided pick list that reserves stock as it goes and becomes a delivery note when complete.

-

### Item substitutes and alternatives
item-substitutes
Set backup items once at the item level so the system already knows the ranked alternative when your first choice runs short, including for manufacturing bills of materials.

-

### Install ERPClaw on the Hermes Agent runtime
install-hermes
Experimental Hermes support via a GitHub tap: the two-command install, why the full skill identifier is mandatory, and the manual clone alternative.

-

### Subcontract production to an outside manufacturer
subcontracting
Send raw materials out, receive finished goods back, and let the books roll material and subcontracting costs into the finished item automatically.

-

### Contacts and companies
contacts-and-companies
Keep people and the businesses they work for as real records, link them, merge duplicates safely, and turn a contact into a paying customer in one step.

-

### Sales tasks and follow-ups
sales-tasks
Track follow-ups with due dates and priorities, linked to the deal, contact, or customer they belong to, so nothing promised in a call dies in a notebook.

-

### Sales pipelines with your own stages
sales-pipelines
Build deal stages that match how you actually sell, not a fixed list, and see every open opportunity grouped by where it really stands.

-

### Saved views and search across Sales
saved-views-and-search
Save a filtered list you actually reuse, share it with the team, and search across every sales record at once instead of remembering where something lives.

-

### Bring your sales data in, and take it out, as CSV
crm-import-export
Import leads, contacts, companies, and deals from a spreadsheet with explicit duplicate handling, and export any list back out. Your data stays yours.

---

## Accounting basics in ERPClaw
URL: https://www.erpclaw.ai/docs/core/accounting/

# Accounting basics in ERPClaw
How the chart of accounts, journal entries, and GL posting work in ERPClaw. Double-entry, immutable submitted entries, the 12-step validation pipeline.
ERPClaw is a double-entry accounting system at its core. Every business document (invoice, payment, payroll run, stock transfer, payout) eventually produces general-ledger entries. This page covers the data model and the rules.

## Chart of accounts

The default install seeds a US-style chart of accounts with 94 accounts grouped by type:

- Assets: 1000-1999 (cash, AR, inventory, fixed assets)

- Liabilities: 2000-2999 (AP, accrued expenses, payroll liabilities, deferred revenue)

- Equity: 3000-3999 (retained earnings, common stock, owner draws)

- Revenue: 4000-4999 (sales, service, other income)

- Expenses: 5000-9999 (COGS, payroll, rent, utilities, depreciation, etc.)

You can customise the chart at any time.

“Add a Marketing expense account, number 5500.”

“Rename it to Paid Marketing.”

The chart supports parent-child hierarchy (so subtotals roll up cleanly) and multi-company (each company has its own set of accounts). Account numbers are unique per company.

## Journal entries

A journal entry is a set of debits and credits that must balance. Total debits must equal total credits.

“Post 1,000 dollars of cash against sales, dated 15 April.”

Submitting runs the 12-step validation pipeline (see below). On success the entry is locked. There is no `update-journal-entry` after submit. To correct a mistake, post a reversing entry. Immutability after submit is the rule.

## The 12-step validation pipeline

Every GL posting passes through this pipeline before it commits. Any failure rolls back the entire transaction; partial writes are impossible.

- Posting date is in an open accounting period

- Company is active and not in lockdown

- All referenced accounts exist

- All accounts allow the posting type (e.g. expense accounts cannot receive a credit-only posting in some configurations)

- Debit total equals credit total exactly (Decimal arithmetic, no float comparison)

- Currency is the company’s functional currency or has a valid exchange rate row

- Cost center references resolve

- Project references resolve

- Tax codes resolve

- Reference document exists if specified

- No conflicting concurrent transaction (FK and uniqueness checks)

- Final balance check after the write (running totals stay consistent)

The pipeline runs inside a single SQLite transaction with `BEGIN ... COMMIT`. If any step fails, the transaction rolls back and you get a clean JSON error.

## How invoices post to GL

When you submit a sales invoice, the GL posting is automatic. A $1,000 invoice for one item in inventory posts like this.

| Account
| Debit
| Credit

| Accounts Receivable
| $1,000.00
|

| Sales Revenue
|
| $1,000.00

| Cost of Goods Sold
| $400.00
|

| Inventory
|
| $400.00

ERPClaw handles the COGS posting automatically based on the item’s costing method (FIFO or weighted average). The invoice voucher links to the four GL entries, so audit trail is one click.

You can name products the way you would say them out loud. Ask to “invoice 20 Brake Pad Sets” and ERPClaw resolves the loose, plural phrasing to the stored product (“Brake Pad Set”), so you do not have to match the exact name on file.

When the customer pays, the second pair of entries follows.

| Account
| Debit
| Credit

| Cash
| $1,000.00
|

| Accounts Receivable
|
| $1,000.00

The AR balance reduces. The trial balance stays balanced. Recording the payment also marks the invoice paid and clears its outstanding balance, so the invoice no longer shows as owing.

## Closing a period

Periods, typically months or quarters, are closed when you say so.

“Close Q1.”

Once closed, no further postings to that period are allowed. The pipeline’s step 1 catches them. Re-opening is rare and usually an auditor finding.

“Reopen Q1, the auditor found an adjustment.”

The reopen requires an explicit reason and is logged.

## Multi-currency

ERPClaw supports multi-currency. Each company has a functional currency (the currency the books are kept in). Foreign-currency invoices and payments stay in the transaction currency end to end; the books reflect the actual currency the transaction occurred in. See the multi-currency plan for the current rollout state.

## Reports

Ask for any of these by name: trial balance, balance sheet, profit and loss, cash flow statement, AR aging, AP aging, general ledger by account, and the journal register.

All reports are exact down to the cent (Decimal arithmetic, no float rounding errors).

## What is next

- Purchasing, purchase orders, goods receipt, vendor invoices, payments.

- Install guide, get to a working ERPClaw in 5 minutes.

- Stripe docs, payment integration that posts directly to your GL.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-account name="Marketing" parent="Expenses" account_number=5500
update-account id=<id> name="Paid Marketing"
list-accounts
add-journal-entry posting_date=2026-04-15 entries='[
{"account": "Cash", "debit": 1000.00},
{"account": "Sales", "credit": 1000.00}
]'
submit-journal-entry id=<id>
close-accounting-period period="2026-Q1" reason="Q1 close"
reopen-accounting-period period="2026-Q1" reason="auditor adjustment"
trial-balance / balance-sheet / profit-and-loss / cash-flow-statement
ar-aging / ap-aging / gl-by-account / journal-register`

See Core features

---

## Contacts and companies
URL: https://www.erpclaw.ai/docs/core/contacts-and-companies/

# Contacts and companies
Keep people and the businesses they work for as real records, link them, merge duplicates safely, and turn a contact into a paying customer in one step.
Sales lives or dies on knowing who you talked to and where they work. ERPClaw has built-in Sales, so the people you are selling to live in the same system as the orders and invoices they eventually generate. No export to a separate CRM, no re-keying the customer when the deal closes.

Say it the way you would out loud:

“Add Jane Smith from Acme Corp, she is their operations head, email jane at acme dot com.”

## People and businesses are separate records

A contact is a person. A company is a business. Jane can move from Acme to a new employer without losing her history, and Acme can have five contacts without you typing its details five times. Link a person to their company with a role, and mark who the primary contact is.

“Link Jane to Acme Corp as their head of operations, she is our main contact there.”

Emails are checked for duplicates as you add people, and company domains the same way, so “acme.com” cannot quietly become two different companies with two different histories.

## Find them again

“Show me everyone at Acme Corp.”

“Which of our contacts are still just leads?”

Lists filter by company, by lifecycle stage, or by a search across names. For a search across everything at once, including deals and customers, see saved views and search.

## Merge the duplicates you find anyway

Every contact list grows duplicates: one from a trade show import, one typed by hand. Merging keeps the record you choose, moves everything the duplicate was attached to over to it, and does the whole thing in one step so a half-merged contact can never exist.

“Jane Smith and J. Smith are the same person, merge them, keep the first one.”

## The step that pays for all of it

When a contact becomes a buyer, promote them:

“Jane just signed. Make her a customer.”

That one request creates the real customer record your quotes, orders, and invoices post against, carried over from the contact you already maintained. If anything goes wrong mid-way, the whole step is rolled back rather than left half-done. This is the point of built-in Sales: the pipeline ends in the same books the rest of the business runs on.

## Common questions

Do I need a separate CRM subscription for this?
No. Contacts, companies, deals, and tasks are part of ERPClaw. Free forever, self-hosted, same install.

What happens to a contact when I remove them?
Removal is soft. The record is retired rather than destroyed, and its links are cleaned up, so history stays intact for anything that referenced them.

Can two contacts share an email address?
Not within the same company records. Email is checked without caring about capitalisation, which is how most duplicates sneak in.

What does merging actually move?
Everything attached to the duplicate: links, roles, and references are reassigned to the record you keep, in one atomic step.

Does promoting a contact copy them or move them?
It creates the customer record from the contact and keeps the contact linked, so the sales history and the billing history stay connected.

Can I bring in my existing contact list?
Yes, from a spreadsheet. See CSV import and export.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-crm-contact --name "Jane Smith" --email jane@example.com --job-title "Head of Operations"
add-crm-company --name "Acme Corp" --domain acme.com
link-contact-to-company --crm-contact-id <id> --crm-company-id <id> --role-title "Head of Operations" --is-primary
list-crm-contacts --crm-company-id <id>
merge-crm-contacts --primary-contact-id <keep> --duplicate-contact-id <merge-away>
promote-contact-to-customer --crm-contact-id <id>
update-crm-contact / get-crm-contact / remove-crm-contact
update-crm-company / get-crm-company / list-crm-companies`
Contact email and company domain are unique per company, checked case-insensitively.

See Core features

---

## Bring your sales data in, and take it out, as CSV
URL: https://www.erpclaw.ai/docs/core/crm-import-export/

# Bring your sales data in, and take it out, as CSV
Import leads, contacts, companies, and deals from a spreadsheet with explicit duplicate handling, and export any list back out. Your data stays yours.
Your leads live in a spreadsheet today, or in a CRM you are leaving. ERPClaw’s built-in Sales takes them in from CSV, the format everything can produce, and gives them back the same way. No import wizard fees, no export ransom.

Say it the way you would out loud:

“Import the leads from the trade show spreadsheet. If someone is already in the system, skip them.”

## Importing: you decide what a duplicate means

Every import makes you say, up front, what happens when a row matches an existing record: skip it, update the existing record with the row, or stop the whole import. There is no silent default, because the wrong guess on that question is how a clean contact list becomes a merge project.

Duplicates are matched the way humans mean it: people by email, companies by web domain, neither caring about capitalisation, the same rules that guard contacts and companies day to day. Deals have no natural duplicate key, so every deal row imports as its own deal, landing in your pipeline at whatever stage the row names.

An import is all or nothing. If a row fails, or you chose “stop on duplicates” and one appears, the entire import rolls back and the system is exactly as it was. Money columns and emails are checked before anything is written, not after.

## Exporting: the door out is always open

“Export every qualified lead to a file for the mailing house.”

Exports write a fresh CSV of leads, deals, contacts, or companies, with simple filters for status or lifecycle, and your custom fields included on request. The file opens correctly in Excel, including names with accents.

A fair test of any system is whether an export, imported straight back with skip-duplicates, changes nothing. Here that round trip is a no-op by design.

## Common questions

What file format do you take?
CSV, one record type per file. Point the import at the file path and state your duplicate rule.

What happens if row 400 of 500 is broken?
Nothing is imported. The import runs as one transaction and rolls back entirely, so you fix the row and run it again rather than reconciling a half-import.

How are duplicates detected?
People by email, companies by domain, both case-insensitive within your company. Deals always insert, since two deals can legitimately share a name.

Can I import from my old CRM?
If it exports CSV, yes. Export each record type from the old system, import them here, and state your duplicate rule per file.

Do my custom fields survive export?
Yes, ask for them and they are appended as extra columns.

Is my data locked in?
No. Export is a first-class action on every one of these record types, and the underlying database is on your own machine to begin with. Free forever, self-hosted.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`import-leads --file leads.csv --on-duplicate skip
import-crm-contacts --file contacts.csv --on-duplicate update
import-crm-companies --file companies.csv --on-duplicate fail
import-opportunities --file deals.csv --on-duplicate skip
export-leads --output qualified-leads.csv --status qualified
export-crm-contacts --output contacts.csv --include-udfs
export-opportunities --output deals.csv
export-crm-companies --output companies.csv`
`--on-duplicate` is required on every import, with skip, update, or fail as the choices. Imports run in a single transaction. Exports overwrite their output file and write UTF-8 with a byte-order mark so spreadsheets open them cleanly.

See Core features

---

## Import a bank statement
URL: https://www.erpclaw.ai/docs/core/import-bank-statement/

# Import a bank statement
Upload an OFX, CAMT.053, MT940, or BAI2 statement, auto-match the lines to open invoices and payments, review what is left, and re-import safely without doubling up.
You can hand ERPClaw a bank statement file and it matches the lines to invoices and payments you already have on the books (the shorter overview is here). Say it the way you would out loud:

“Import my January bank statement and match it to open invoices.”

## The formats you can upload

ERPClaw reads the four statement formats banks actually export.

- OFX (Open Financial Exchange), the common download from US and Canadian banks

- CAMT.053, the ISO 20022 statement most European banks issue

- MT940, the long-standing SWIFT statement format

- BAI2, the cash-management format from US corporate banking

Pick whichever your bank gives you. There is nothing extra to install for any of them, and ERPClaw detects the format on its own unless you name it.

## Step 1, import the file

“Import my January bank statement for the checking account.”

Point ERPClaw at the file and name the bank account the statement belongs to. That is all it needs; the format is worked out for you unless you would rather be explicit.

The bank account is resolved by name within your company. If you name an account that does not exist, ERPClaw stops and tells you, rather than guessing a different account. It never quietly imports a statement into the wrong place. Each transaction on the statement becomes a line on the imported record, ready to match.

## Step 2, auto-match to open invoices and payments

“Match those statement lines against our open invoices and payments.”

The matching engine runs your rules over the imported lines and links each one it can place to an open invoice or payment. A deposit that clears an outstanding customer invoice gets paired with that invoice; a withdrawal that settles a bill gets paired with that payment. The lines it is confident about are matched in one pass.

## Teach it your patterns with match rules

Recurring lines can match themselves once you have told ERPClaw what they are.

“Anything with STRIPE in the description is a Stripe payout, map it to that account.”

A rule looks at one field on the line (the description, the counterparty name, the reference, or the amount), tests it (equals, contains, a regular expression, or an amount range), and on a hit routes the line to an account, a vendor, or a customer, or tells the engine to leave it alone. Give rules a priority to control which one wins when more than one could apply, and ask to see the rules you have whenever you want to review them.

## Step 3, review the unmatched lines

The engine leaves anything it cannot place with confidence for you to look at.

“What is still unmatched on that statement?”

“That one is a payment to Acme Steel, match it to them.”

Changed your mind about a match? Clear it and the line goes back into the unmatched pile, ready to be placed somewhere else.

## Re-import safety, no duplicates

Re-importing the same file is safe. Every line carries its own identifier from the bank, and any line already on the books is skipped on the way in. So if a download was interrupted, or you are not sure whether January already went through, you can run the import again without doubling up the statement.

## Reconcile and close out

When the lines are matched, check where the account stands.

“How does the checking account reconcile as of the end of January?”

You can ask for the picture at any cutoff date, list the statements imported so far, open any one of them, and archive a statement you are finished with.

## What is next

- Reports by department, project, or any dimension splits your P&L the way you run the business.

- Accounting basics covers the chart of accounts, journals, and how postings flow.

- Purchasing walks the purchase order, receipt, vendor invoice, and payment loop.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`integration-import-bank-statement --file ~/statements/checking-january.ofx --bank-account-name "Checking"
integration-auto-match-bank-statement --statement-id <id>
integration-add-bank-match-rule --name "Stripe payouts" --match-field description --match-operator contains --match-value "STRIPE" --target-action map_to_account
integration-list-bank-match-rules
integration-unmatched-bank-lines --statement-id <id>
integration-manual-match-bank-line --line-id <id> --target-action map_to_vendor --target-id <vendor-id>
integration-clear-bank-line-match --line-id <id>
integration-bank-reconciliation-summary --bank-account-name "Checking" [--as-of <date>]
integration-list-bank-statements
integration-get-bank-statement --statement-id <id>
integration-archive-bank-statement --statement-id <id>`
Match-rule fields are `description`, `counterparty_name`, `reference`, or `amount`. Operators are `equals`, `contains`, `regex`, or `amount_range` written as `min:max`. Target actions are `map_to_account`, `map_to_vendor`, `map_to_customer`, or `ignore`. Import accepts `--format` (`ofx`, `camt053`, `mt940`, `bai2`) when you want to be explicit.

See Core features

---

## Install ERPClaw on the Hermes Agent runtime
URL: https://www.erpclaw.ai/docs/core/install-hermes/

# Install ERPClaw on the Hermes Agent runtime
Experimental Hermes support via a GitHub tap: the two-command install, why the full skill identifier is mandatory, and the manual clone alternative.
Runs on OpenClaw (primary). Experimental support for the Hermes Agent runtime via a GitHub tap.

That sentence is the support model, and it is worth reading precisely. OpenClaw is the runtime ERPClaw is built and tested against. Hermes works, and it is experimental, which means the tap install path is supported but the runtime is not the primary target.

This page is for developers and early adopters comfortable with that distinction.

## Install

Two commands, in this order:

`hermes skills tap add avansaber/hermes-skills
hermes skills install avansaber/hermes-skills/skills/erpclaw --force`
Use the full identifier exactly as written. `avansaber/hermes-skills/skills/erpclaw` is mandatory. Shorter forms resolve silently to a stale marketplace copy, and the failure is quiet: the install appears to succeed and you end up running old code while reading current documentation. If you take one thing from this page, take that.

## Why the install needs `--force`

`--force` acknowledges the caution rating the skill audit assigns. That rating is inherent to what ERPClaw is: an ERP that runs local commands against a local database. A tool that posts journal entries and migrates schemas on your machine is correctly flagged as something to install deliberately rather than by accident.

The flag is not a workaround for a broken package. It is an acknowledgement, and requiring it is the right behaviour from the runtime.

## Set the install root

`export ERPCLAW_HOME=~/.hermes/erpclaw-home`
`ERPCLAW_HOME` sets the install root, and the library, the install state, and the default SQLite database all resolve underneath it. Leave it unset or blank and it defaults to `~/.openclaw/erpclaw`, byte identical to the OpenClaw layout.

Set it explicitly when you want the Hermes install kept separate from an existing OpenClaw install on the same machine. Leave it alone if you want both runtimes pointed at the same books.

## The manual alternative

If you would rather not pass `--force`, clone the tap and install from the local path instead. The current copy of that procedure lives in the README of the `avansaber/hermes-skills` tap, so it stays correct as the tap evolves rather than drifting on this page.

## After install

The post-install step initialises the database, and from there the usual first run applies: set up the company, then work in plain language. The core install guide covers the setup path in full, and it is the same once the skill is in place regardless of which runtime loaded it. From there, accounting basics is the natural next page.

## Common questions

Is Hermes supported or not?
Experimentally, via the GitHub tap. OpenClaw is the primary runtime. Both parts of that sentence are true and neither should be read as the whole answer.

What happens if I use a shorter skill identifier?
It resolves to a stale marketplace copy without telling you. Always install with the full `avansaber/hermes-skills/skills/erpclaw` path.

Can I skip `--force`?
Yes, by cloning the tap and installing from the local path. The tap README carries that procedure.

Where does ERPClaw put its files under Hermes?
Wherever `ERPCLAW_HOME` points. Unset or blank means `~/.openclaw/erpclaw`, which is the same layout OpenClaw uses.

Can one machine run both runtimes?
Yes. Point them at the same `ERPCLAW_HOME` to share one set of books, or at different roots to keep them separate.

Is the feature set different on Hermes?
The skill is the same skill. What differs is the level of support behind the runtime it is loaded into.

See Core features

---

## Install ERPClaw in 5 minutes
URL: https://www.erpclaw.ai/docs/core/install/

# Install ERPClaw in 5 minutes
Step-by-step install guide for the ERPClaw core. Covers OpenClaw + ClawHub install, database init, first commands, and verification.
ERPClaw installs as one OpenClaw skill. Every module, every action, and every table shares one SQLite database at `~/.openclaw/erpclaw/data.sqlite`. PostgreSQL is supported as an alternative backend via the same code (PyPika handles the dialect translation). Five minutes from zero to a working ERP.

## Prerequisites

You need Python 3.10 or newer (most macOS and Linux installs already have this). You do not need PostgreSQL, Redis, Docker, or a cloud account. Everything runs locally.

To verify Python:

`python3 --version`
If you see anything `3.10` or higher, you are good.

## Step 1, install OpenClaw

OpenClaw is the agent runtime that hosts ERPClaw. The installer is a single shell command:

`curl -fsSL https://openclaw.ai/install.sh | bash`
The installer drops `openclaw` and `clawhub` binaries into `~/.npm-global/bin/` and prints completion instructions. Add the bin directory to your `PATH` if it is not already there. Most users do not need to change anything.

## Step 2, install ERPClaw via ClawHub

`clawhub install erpclaw`
ClawHub fetches the latest ERPClaw release (version 3.5.0 as of 2026-04-26) and unpacks it into `~/.openclaw/workspace/skills/erpclaw/`. The install also brings in the shared `erpclaw_lib` library and the module registry.

The first install pulls about 5 MB of code. No images, no docker layers, no node_modules.

## Step 3, initialize the database

`python3 ~/.openclaw/workspace/skills/erpclaw/init_db.py`
This creates the SQLite database at `~/.openclaw/erpclaw/data.sqlite` with 191 core tables. It also seeds the US chart of accounts (94 accounts pre-built), default tax templates, and a single demo company called Stark Manufacturing.

If you want to skip the demo data, pass `--no-seed`:

`python3 ~/.openclaw/workspace/skills/erpclaw/init_db.py --no-seed`

## Step 4, verify the install

Open a chat with OpenClaw (or use the CLI directly) and try:

`list-companies`
You should see Stark Manufacturing in the result. Try a couple more:

`list-customers
list-items
trial-balance`
If `trial-balance` returns a balanced sheet, GL is wired correctly. The demo data has 102 GL entries, 33 stock-ledger entries, and a balanced trial balance of $194,389.50.

## Step 5, install verticals as needed

ERPClaw ships 14 industry verticals: retail, restaurant, healthcare, legal, nonprofit, education, real estate, agriculture, automotive, food, hospitality, construction, fleet, logistics. Install only what you need.

`clawhub install retailclaw
clawhub install hospitalityclaw`
Each vertical adds its own tables on top of the shared core. You never have to migrate data between modules; they all read and write to the same database.

## Switching to PostgreSQL

If you outgrow SQLite (typically at 100+ concurrent users or strict ACID requirements), switch to PostgreSQL by setting one environment variable:

`export ERPCLAW_DB_URL=postgresql://user:pass@host:5432/erpclaw
python3 ~/.openclaw/workspace/skills/erpclaw/init_db.py`
The same actions work without any code change. PyPika translates the SQL dialect at query time.

## Where things live

- Code: `~/.openclaw/workspace/skills/erpclaw/`

- Database: `~/.openclaw/erpclaw/data.sqlite`

- Shared library: `~/.openclaw/erpclaw/lib/erpclaw_lib/`

- Vertical modules: `~/.openclaw/erpclaw/modules/<module>/`

## What is next

- Accounting basics covers the chart of accounts, journals, and how postings flow.

- Purchasing walks the purchase order to invoice to payment loop.

- Stripe docs for the payment integration.

- Shopify docs for the ecommerce integration.

Total install time on a modern laptop is under five minutes. If anything fails, check the troubleshooting section or open an issue.

See Core features

---

## Item substitutes and alternatives
URL: https://www.erpclaw.ai/docs/core/item-substitutes/

# Item substitutes and alternatives
Set backup items once at the item level so the system already knows the ranked alternative when your first choice runs short, including for manufacturing bills of materials.
Every operation has substitutions that live in somebody’s head: if the 12mm bolts run out, the 12.5mm ones will do. Writing that down once, at the item level, means the system can answer the question instead of the one person who knows.

## Set an alternative

“If we run out of 12mm bolts, the 12.5mm ones are the first fallback.”

Three things shape the behaviour:

- Priority ranks the alternatives. Lower is preferred, so priority 10 is chosen before priority 20.

- A conversion factor handles the case where the substitution is not one for one.

- Notes record why the substitution is acceptable, which is the part that saves an argument later.

Two rules are worth knowing before you load a list.

- An item cannot be its own alternative. Self-references are rejected.

- The pair is directional. Recording that A substitutes for B does not mean B substitutes for A, and both directions can exist as separate rows if both are genuinely true.

That second rule catches people out, and it is deliberate. Substitutability is often one-way: the premium part can stand in for the standard one, while the reverse would ship a customer the wrong thing.

## Ask what to use

“We need 500 of the 12mm bolts at Main and we are short. What should we use instead?”

This returns the highest-priority active alternative that actually has enough stock at that warehouse, so the answer accounts for reality rather than just preference order. Ties are broken by available quantity, which favours the substitution least likely to run short next.

When nothing qualifies you get a clean empty result rather than an error. No alternative is a legitimate answer, and the caller can act on it.

## Review and remove

“What alternatives do we have on file?”

“Drop the 12.5mm bolt as a substitute, it is discontinued.”

Removing soft-disables the row. The substitution stops being offered and the record of it having existed remains, which matters when someone asks why a build in March used a different part.

## Manufacturing inherits these

Bills of materials get this behaviour without being configured twice. On the fulfilment side the same list is what a short pick list reaches for, and what you consult before placing a stock reservation you may not be able to honour. When a BOM line has no substitutes of its own, it inherits the item-global alternatives set here. Define the substitution once at the item level and it applies to production as well as to fulfilment, unless a specific BOM line overrides it.

## Common questions

Does a lower priority number mean less preferred?
No, the opposite. Lower priority is chosen first. Priority 10 beats priority 20.

Why is the relationship directional?
Because substitutability usually is. A higher-specification part can often stand in for a standard one while the reverse would be wrong. If both directions are genuinely valid, record both.

What if the substitute is not a one-for-one swap?
Set `--conversion-factor` when you create the alternative, so quantities translate correctly rather than being assumed equal.

Does the system pick a substitute automatically?
`get-best-alternative-for-item` gives you the ranked answer with stock taken into account. What you do with it stays a decision, which is the right split for something that changes what a customer receives.

What happens when no alternative has enough stock?
You get an empty result, not an error. That is a real answer and the caller can act on it.

Do I have to configure substitutes separately for manufacturing?
No. A BOM line with no substitutes of its own inherits the item-global alternatives.

Does removing an alternative delete the history?
No, it soft-disables the row so past decisions remain explainable.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-item-alternative --item BOLT-12MM --alternative BOLT-125MM --priority 10
get-best-alternative-for-item --item BOLT-12MM --required-qty 500 --warehouse "Main"
list-item-alternatives
remove-item-alternative --id <alternative-id>`

See Core features

---

## Purchasing in ERPClaw
URL: https://www.erpclaw.ai/docs/core/purchasing/

# Purchasing in ERPClaw
How purchase orders, goods receipts, vendor invoices, and payments flow in ERPClaw. Three-way matching, GL postings at each step, supplier setup.
The purchasing flow in ERPClaw is procure to pay: PO, goods receipt, vendor invoice, payment. Three-way matching happens before the invoice is approved. Every step writes the right GL entries automatically.

## The four-step flow

- Create a purchase order with the items and quantities you want to buy.

- Receive the goods when they arrive at your warehouse.

- Match the vendor invoice against the PO and receipt (three-way matching).

- Pay the vendor when the invoice is due.

Each step is a separate document in ERPClaw with its own GL postings.

## Step 1, supplier setup

Before you can buy from a supplier, the supplier must exist.

“Add Acme Steel as a supplier, billing at billing@example.com, Net 30 terms.”

Supplier records hold contact info, default payment terms, default currency, tax IDs, and bank details if you want to ACH payments. You can store a dedicated email and phone on each supplier (and on each customer) when you add or update the record, so the people you bill and the people you buy from each keep their own contact details.

## Step 2, the purchase order

A purchase order is an offer to buy. It does not post to GL on its own (no goods received, no invoice yet). It just commits you to the vendor in writing.

“Raise a purchase order to Acme Steel for 5 steel beams at 200 each and 100 quarter-inch bolts at 50 cents, for delivery by the end of April.”

The submit step locks the PO. To cancel, post a cancellation; to amend, create a new PO and reference the old one.

You do not have to match a product name exactly. Refer to an item loosely or in the plural (“5 Steel Beams”) and ERPClaw resolves it to the stored product (“Steel Beam, 10ft”), so everyday phrasing works without you hunting for the exact name on file.

## Step 3, goods receipt

“The steel beams arrived, receive them into the main warehouse.”

Logging the receipt is one step, and it is where inventory actually moves.

The receipt posts to GL for you.

| Account
| Debit
| Credit

| Inventory
| $1,050.00
|

| Goods Received Not Invoiced
|
| $1,050.00

The “Goods Received Not Invoiced” (GRNI) account is a temporary holding account that clears when the vendor invoice arrives. This is a clearing-account pattern and it is the right way to handle the timing gap between receipt and invoice.

When you receive stock you bought, ERPClaw values it from the unit cost on the purchase order or bill, so your inventory is costed and booked correctly with no $0 surprises. If a receipt arrives with no cost to work from, ERPClaw refuses it rather than booking the stock at zero.

## Step 4, vendor invoice

“Acme sent their invoice for that order, match it against the PO and the receipt.” with three-way matching

Matching the invoice against the PO and the receipt is the three-way match.

Three-way matching compares three things.

- PO quantity vs receipt quantity (did we receive what we ordered?)

- Receipt quantity vs invoice quantity (is the vendor billing for what arrived?)

- PO unit price vs invoice unit price (is the vendor honoring the agreed price?)

If all three match within tolerance, the invoice can be submitted. If not, you get a mismatch report and the invoice stays in draft until reconciled.

On submit the GL posting follows.

| Account
| Debit
| Credit

| Goods Received Not Invoiced
| $1,050.00
|

| Accounts Payable
|
| $1,050.00

The GRNI account clears. AP now owes the vendor.

## Step 5, payment

Pay the vendor when the invoice is due.

The GL posting is automatic.

| Account
| Debit
| Credit

| Accounts Payable
| $1,050.00
|

| Cash
|
| $1,050.00

AP balance now zero for that invoice. Cash account drops. Recording the payment also marks the vendor bill paid and clears its outstanding balance, so the bill no longer shows as owing.

## When goods arrive without a PO

Sometimes a one-off purchase has no PO, so you can skip that step entirely.

“Bill 800 dollars of office chairs from Office Depot to furniture expense.”

The invoice posts directly to the expense account and AP. No three-way matching because there is nothing to match against.

## Returns and credit notes

To return goods to a vendor, post a credit note against the original invoice.

“Return one steel beam to Acme and credit the invoice.”

The credit note reverses the partial GL impact and reduces AP by the credit amount.

## Reports

Ask for any of these by name: open POs, pending receipts, unmatched invoices, and supplier balances.

- AP aging (`ap-aging`)

- Vendor balance (`vendor-balance` )

- Three-way match exceptions (`three-way-match-exceptions`)

- Purchase register (`purchase-register`)

## What is next

- Accounting basics, chart of accounts, journals, GL postings.

- Install guide, get to a working ERPClaw in 5 minutes.

- Three-way matching, plain English, the AP staff guide.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-supplier name="Acme Steel" email=billing@example.com payment_terms="Net 30"
list-suppliers
add-purchase-order supplier="Acme Steel" delivery_date=2026-04-30 items='[
{"item": "Steel Beam, 10ft", "quantity": 5, "rate": 200.00},
{"item": "Bolts, 1/4 inch", "quantity": 100, "rate": 0.50}
]'
submit-purchase-order id=<id>
create-purchase-receipt purchase_order_id=<po-id> received='[
{"item": "Steel Beam, 10ft", "quantity": 5},
{"item": "Bolts, 1/4 inch", "quantity": 100}
]'
submit-purchase-receipt id=<receipt-id>
create-purchase-invoice purchase_order_id=<po-id> purchase_receipt_id=<receipt-id> invoice_amount=1050.00
add-payment-entry payment_type=pay party_type=supplier party_id=<supplier-id> paid_amount=1050.00 reference_number=ACH-12345
submit-payment-entry id=<payment-id>
create-purchase-invoice supplier="Office Depot" line_items='[{"description": "office chairs", "amount": 800.00, "expense_account": "Office Supplies"}]'
submit-purchase-invoice id=<id>
create-purchase-credit-note purchase_invoice_id=<id> items='[{"item": "Steel Beam, 10ft", "quantity": 1, "rate": 200.00}]'
submit-purchase-credit-note id=<cn-id>`

See Core features

---

## Putaway rules and pick lists
URL: https://www.erpclaw.ai/docs/core/putaway-and-pick-lists/

# Putaway rules and pick lists
Route incoming goods to the right warehouse automatically, and fill orders from a guided pick list that reserves stock as it goes and becomes a delivery note when complete.
Two ends of the same day: goods arrive and need to land in the right place, and orders go out and need to be picked without anyone guessing. Putaway rules handle the first, pick lists handle the second, and both are deterministic so two people running the same receipt get the same answer.

## Putaway: decide where received stock goes

A putaway rule says “this item, or this group of items, belongs in this warehouse.” You write the rules once and receiving stops being a judgment call.

“Blue chairs always go to Main. Everything else in Furniture goes to Overflow.”

Matching is resolved in a fixed order, which is what makes it predictable.

- An item match (`--match-item`) beats an item-group match (`--match-item-group`).

- Within the same kind of match, lower `--priority` wins.

So the specific rule always beats the general one, and you can layer a catch-all group rule underneath item-level exceptions without them fighting.

Review and adjust in the same way. “Show me the putaway rules” gets you the list, and “move the furniture rule ahead of the chair rule” changes the priority. Turning a rule off soft-disables it, so it stops applying and stays on the record.

Deleting soft-disables the rule rather than erasing it, so a rule you turn off is still auditable later.

## Apply putaway to a receipt

“Where does this delivery go?”

This computes the routing for a `material_receipt` and tells you where each line lands. It is deterministic: the same receipt and the same rules produce the same routing every time, which matters when someone asks six weeks later why a pallet ended up where it did.

Note the scope. Putaway here is warehouse-level routing, not bin-level slotting inside a building.

## Pick lists: fill an order without guesswork

A pick list turns an open sales order into a list someone can walk, holds the stock while they walk it, and becomes a delivery note at the end.

“Make a pick list for order 1042.”

That drafts a pick from the open lines on the order. Where the pick is not order-driven you can add lines yourself, item by item.

## Submit, pick, complete

“Release that pick list to the floor.”

Submitting places hard reservations for the quantities on the list. From this point the stock is spoken for, and an issue that would breach it is refused. That is the whole point: the picker walks out to the aisle knowing the quantity is still there.

Then record what was actually picked, line by line. In practice this is the picker reporting back: “picked all 40 blue chairs.”

Actuals matter more than intentions here. When every line is fully picked the list moves to `picked`, and short picks stay visible rather than being rounded away.

Completing consumes the reservations and generates the delivery note, so the paperwork follows the physical movement instead of being typed in afterwards.

If the pick is abandoned, cancel it.

Cancelling releases the reservations, and the stock is immediately available to everyone else again.

## Common questions

Does putaway work at bin level?
No. It routes to a warehouse. Bin-level slotting inside a building is a different problem and this feature does not claim to solve it.

What if two rules could match the same item?
The order is fixed: item match beats item-group match, then lower priority wins. There is no ambiguity to resolve at receipt time.

Does deleting a rule lose the history?
No, it soft-disables. The rule stops applying and stays on the record.

Why does submitting a pick list reserve stock rather than picking it?
Because the gap between drafting a pick and walking it is where double-promising happens. The reservation closes that gap. See stock reservations for the mechanism, and item substitutes for the ranked fallback when a line is short.

What happens to reservations if a pick is only partly filled?
The list records the actual picked quantity and stays open. Completing it consumes the reservations behind the quantities actually picked; cancelling releases everything.

Do I have to start from a sales order?
No. `create-pick-list --from-sales-order` is the common path because most picks are order-driven, but you can build a list line by line with `add-pick-list-item`.

Does completing a pick create the delivery paperwork?
Yes. `complete-pick-list` generates the delivery note as part of completion.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-putaway-rule --match-item CHAIR-BLUE --warehouse "Main"
add-putaway-rule --match-item-group "Furniture" --warehouse "Overflow" --priority 20
list-putaway-rules
update-putaway-rule --id <rule-id> --priority 10
delete-putaway-rule --id <rule-id>
apply-putaway-on-receipt --stock-entry <stock-entry-id>
create-pick-list --from-sales-order SO-1042
add-pick-list-item --pick-list <pick-list-id> --item CHAIR-BLUE --qty 40
submit-pick-list --id <pick-list-id>
mark-picked --pick-list <pick-list-id> --item CHAIR-BLUE --picked-qty 40
complete-pick-list --id <pick-list-id>
cancel-pick-list --id <pick-list-id>`

See Core features

---

## Reports by department, project, or any dimension
URL: https://www.erpclaw.ai/docs/core/reports-by-dimension/

# Reports by department, project, or any dimension
Tag entries with a department, project, or any dimension you define, then ask for a profit and loss split by it, with untagged amounts shown separately.
ERPClaw can split your numbers by department, project, cost center, location, fund, or any label you define. The trick is two steps: tag an entry when you book it, then ask for the report.

“Show me this month’s P&L broken down by department.”

That is all it takes once your entries are tagged. For the shorter overview of
what this answers, see dimensional reporting.

## Tag entries when you book them

When a posting belongs to a department or a project, say so at the time you book it.

“Book 120 dollars of office supplies against Engineering.”

The same tagging works when you raise an invoice or record a payment. That tag is what makes the entry show up under the right heading later. You do not add a “cost center” column or keep the attribution in a spreadsheet on the side; the tag travels with the posting for good.

## Ask for P&L by a dimension

Once entries are tagged, ask for the breakdown.

“Profit and loss for June, grouped by department.”

You get revenue, expenses, and net for each department, across your income and expense accounts. Ask without the grouping and you get the single company-wide profit and loss exactly as before. You group by one dimension at a time, so pick the lens you want for this report: department, project, or whichever you defined.

## The (untagged) bucket

Anything that is not tagged with the dimension you asked for is never dropped from the report. It lands in an explicit `(untagged)` bucket of its own. That means the parts always add back up to the company total, and you can see at a glance how much is still unattributed and worth going back to tag.

## Define your own dimensions

Out of the box you can tag by project, department, and cost center. Add the ones your business actually runs on, like region, fund, campaign, or store.

“Add a region dimension, and label it Sales region.”

“We do not split by region any more, retire it.”

Retiring one is blocked while recent entries still reference it, so you cannot orphan live data by accident. And if you ask to group by a dimension that was never registered, or one you have already retired, ERPClaw stops and tells you which dimensions exist instead of handing you a confusing empty split.

## Beyond the P&L

Profit and loss is the common ask, but the same tags drive more.

- “Break the trial balance out by project and department” gives you the two-dimension version across the whole trial balance.

- “Show me balances by project” gives one dimension’s balances on their own.

- The trial balance, balance sheet, cash flow statement, and general ledger can all be narrowed to a single department or project. On profit and loss the filter applies first and the grouping happens after, so you can scope to one project and still split it by department.

## What is next

- Import a bank statement matches a statement file to your open invoices and payments.

- Accounting basics covers the chart of accounts, journals, and how postings flow.

- Purchasing walks the purchase order, receipt, vendor invoice, and payment loop.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-journal-entry posting_date=2026-06-12 entries='[
{"account": "Office Supplies", "debit": 120.00},
{"account": "Cash", "credit": 120.00}
]' --dimension-key department --dimension-value Engineering
profit-and-loss --group-by department --from-date 2026-06-01 --to-date 2026-06-30
multi-dim-trial-balance --group-by "project,department"
dimension-balance-report --dimension project
add-dimension --name region
list-dimensions
update-dimension --name region --label "Sales region"
deactivate-dimension --name region`
`trial-balance`, `balance-sheet`, `cash-flow`, and `general-ledger` all accept `--dimension-key` and `--dimension-value` filters. The same `--dimension-key` and `--dimension-value` pair works on invoices and payments, not just journal entries.

See Core features

---

## Sales pipelines with your own stages
URL: https://www.erpclaw.ai/docs/core/sales-pipelines/

# Sales pipelines with your own stages
Build deal stages that match how you actually sell, not a fixed list, and see every open opportunity grouped by where it really stands.
Every business sells differently. An agency moves from brief to proposal to retainer. A manufacturer moves from enquiry to sample to trial order. A fixed set of deal stages forces both of them to lie to their own pipeline report. ERPClaw’s built-in Sales lets you define the stages your deals actually pass through, and the pipeline report groups by those.

Say it the way you would out loud:

“Our deals go enquiry, sample sent, trial order, contract. Set the pipeline up that way.”

## Start with the default, change it when it lies

A standard sales pipeline is seeded on day one, so the report works before you have configured anything: new, contacted, qualified, proposal, negotiation, and the two endings. When those names stop matching reality, build your own:

“Add a stage called Sample Sent, after Contacted, and give it a win probability.”

Stages carry an order and a win probability, and exactly one winning and one losing end point each, which the system enforces rather than trusts. Inserting a stage mid-pipeline shifts the rest down for you.

## Move deals as they move

“The Acme deal just got to Trial Order.”

A deal advances stage by stage along its own pipeline. Moving a deal to a stage from a different pipeline is refused, because a deal that is simultaneously in two selling motions is a report that means nothing. Won and lost are terminal: once a deal ends, it is frozen, and the record of how it ended stays put.

## Ask where the money is

“Show me the pipeline.”

“What is sitting in Negotiation right now?”

The pipeline report groups every open deal by its stage, with expected revenue, so the answer to “what might close this quarter” is a question rather than a spreadsheet afternoon. Deals from the era before you customised anything still report correctly. The people behind those deals live in contacts and companies, and the follow-ups that move them along are sales tasks.

## Common questions

Do I have to design a pipeline before using Sales?
No. A sensible default pipeline is seeded at install. Customise when your process outgrows it.

Can I run more than one pipeline?
Yes. A product line with a fast self-serve motion and an enterprise motion with committees can each have their own stages, and each deal belongs to one.

What stops two people creating conflicting stage orders?
Stage positions are enforced. Two stages cannot occupy the same slot, and each pipeline has exactly one won and one lost ending.

What happens to old deals when I rename or add stages?
Existing deals keep reporting. Legacy stage names still work, so history does not break the day you customise.

Can I reopen a lost deal?
No. Won and lost are frozen endings. If the buyer comes back, that is a new deal, and your win-rate history stays honest.

Where do the probabilities show up?
On the pipeline report, where stage probability against expected revenue is what turns a list of deals into a forecast.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-crm-pipeline --name "Manufacturing Sales" --set-as-default
add-crm-pipeline-stage --pipeline <id> --name "Sample Sent" --order 3 --probability 40 --shift-existing
update-crm-pipeline-stage --id <id> --name "Trial Order"
list-crm-pipelines
list-crm-pipeline-stages --pipeline <id>
set-opportunity-pipeline-stage --opportunity <id> --stage <stage-id>
pipeline-report`
Cross-pipeline stage moves are blocked, terminal stages are unique per pipeline, and `pipeline-report` groups by custom stages where they exist and by legacy stage names where they do not.

See Core features

---

## Sales tasks and follow-ups
URL: https://www.erpclaw.ai/docs/core/sales-tasks/

# Sales tasks and follow-ups
Track follow-ups with due dates and priorities, linked to the deal, contact, or customer they belong to, so nothing promised in a call dies in a notebook.
Deals rarely die because the product lost. They die because the follow-up never happened. ERPClaw’s built-in Sales keeps follow-ups as real tasks with owners, due dates, and priorities, attached to the deal or person they belong to, in the same system as everything else.

Say it the way you would out loud:

“Remind me to send Acme the revised quote by Friday, high priority, and attach it to their deal.”

## A task knows what it is about

Tasks link to the things they concern: a lead, a deal, a customer, or the contacts and companies you maintain. One task can link to several at once, so “call Jane about the Acme renewal” can sit on both Jane and the renewal deal, and turn up when you look at either.

The link is checked when you make it. A task cannot point at a deal that does not exist, which is how orphaned reminders are born in looser systems.

## The morning question

“What is overdue?”

“What is due in the next three days?”

That is the whole workflow. Ask for the overdue list at the start of the day, work it, and ask again tomorrow. You can narrow by owner, by priority, or by what the task is linked to:

“Show me everything open on the Acme deal.”

## Closing the loop

“Done with the quote task, I sent it over this morning.”

Completing a task records when and optionally what happened. A task that is already done refuses to be done twice, so the record stays honest. Cancelled tasks keep their reason, which is worth thirty seconds when you are reconstructing a deal six months later.

## Common questions

How is this different from the activity log?
Activities record what happened: a call was made, an email was sent. Tasks are what should happen next, with a due date and an owner. Both live on the deal, and neither replaces the other.

Can a task belong to more than one thing?
Yes. Link it to the deal and the contact, and it appears on both. Links can be added and removed as the situation changes.

What happens to tasks when a deal closes?
Nothing automatic. Open tasks stay open until completed or cancelled, so a won deal’s onboarding follow-ups survive the win.

Can I assign a task to someone else?
Yes, tasks carry an owner. Filter the list by owner to see one person’s plate.

What counts as overdue?
Anything past its due date and still open. The overdue filter is the fastest honest answer to “what did we drop?”

Do tasks show up in searches?
Yes, alongside contacts, companies, leads, and deals. See saved views and search.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-crm-task --subject "Send revised quote" --due-date 2026-08-21 --priority high --link-to "opportunity:<id>"
list-crm-tasks --overdue
list-crm-tasks --due-within-days 3 --assigned-to <user>
list-crm-tasks --linked-to "opportunity:<id>"
complete-crm-task --crm-task-id <id> --notes "Sent this morning"
cancel-crm-task --crm-task-id <id> --reason "Deal went dark"
link-task-to-entity / unlink-task-from-entity
update-crm-task / get-crm-task`
`--link-to` repeats for multiple links and accepts lead, opportunity, customer, crm_contact, and crm_company targets, checked for existence at link time.

See Core features

---

## Saved views and search across Sales
URL: https://www.erpclaw.ai/docs/core/saved-views-and-search/

# Saved views and search across Sales
Save a filtered list you actually reuse, share it with the team, and search across every sales record at once instead of remembering where something lives.
The lists you look at every week deserve names. “Leads from the trade show that nobody has called”, “deals over ten thousand stuck in proposal”, “customers in Texas”. ERPClaw’s built-in Sales lets you save a filter once and replay it by name, and when you cannot remember where a record lives at all, one search sweeps everything.

Say it the way you would out loud:

“Save a view called Hot Leads: status qualified, sorted by newest first. Share it with the team.”

## A view is a filter with a name

A saved view belongs to one kind of record, such as leads, deals, customers, or the contacts and companies you maintain, and holds the filter, the sort order, and the grouping you chose. A view built on your pipeline stages is how “deals stuck in proposal” becomes one click. Replaying it gives you today’s records through that same lens:

“Show me the Hot Leads view.”

Views are checked when you save them and again when you run them, against the fields that record actually has, so a view cannot silently reference a field that was never there or return something other than what it claims.

## Yours or everyone’s

A view is private to its owner unless shared. Shared views are readable by the whole company, and only the owner can change or delete one. That split matters: the team’s “This Quarter’s Committed” view should not be editable by whoever opened it last.

## When you cannot remember where it lives

“Search everything for Acme.”

One search runs across leads, deals, customers, contacts, and companies at once and comes back ranked, exact matches first, then close ones. The result names which kind of record each hit is, so “Acme” the company, “Acme renewal” the deal, and “jane@example.com” the contact all land in one list.

## Common questions

What is the difference between a saved view and a search?
A search is a question you ask once. A view is a question you ask every week, saved with its filters and sorting so you never rebuild it.

Can views filter on my custom fields?
Yes. Fields you have added to a record type are usable in views on that type, checked with the same care as built-in fields.

Who sees a shared view?
Everyone in the company can use it. Only the owner can edit or delete it.

How current are view results?
Live. A view stores the question, not the answers, so replaying it always reflects today’s records.

Is there a way to misuse filters to pull data out?
Filter values are handled as data, never as code, and filters are validated against an allowlist of fields and comparisons on both save and run.

How short can a search be?
Two characters. Results are ranked and capped, so a broad search stays usable instead of endless.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-crm-saved-view --name "Hot Leads" --entity-type lead --filter-json '{"field":"status","op":"eq","value":"qualified"}' --is-shared
apply-saved-view --view <id>
list-crm-saved-views --entity-type lead
update-crm-saved-view --id <id> --owner-user-id <owner>
delete-crm-saved-view --id <id> --owner-user-id <owner>
global-crm-search --query "Acme"
list-leads --saved-view-id <id>`
Filter JSON supports eq, neq, contains, gt, lt, in, and between, validated against each record type’s field allowlist, with values always bound as parameters.

See Core features

---

## Reserve stock for an order
URL: https://www.erpclaw.ai/docs/core/stock-reservations/

# Reserve stock for an order
Hold a quantity for one order so another order cannot quietly consume it. How reservations are checked, what gets refused, and how to release stock you no longer need to hold.
Promising the same stock twice is one of the easiest ways to lose a customer. A reservation holds a quantity against a specific voucher so nothing else can consume it, and an issue that would break the promise is refused rather than silently allowed.

Say it the way you would out loud: “hold 40 of the blue chairs at the main warehouse for this order.”

## What a reservation actually does

Reservations here are hard, not advisory. Three things follow from that:

- Availability is reduced for everyone else. A new reservation is refused if it would exceed available stock, where available means actual quantity minus everything already actively reserved.

- An issue that would breach an active reservation is blocked. A `material_issue` that would eat into reserved quantity does not go through, so the promise survives contact with the rest of the system.

- Held stock stays visible. The quantity is not hidden or moved; it is on hand and spoken for.

## Reserve a quantity

Ask for it the way you would say it to a colleague:

“Hold 40 blue chairs at the main warehouse for this sales order.”

ERPClaw resolves the item from your loose phrasing and places the hold against that order, for that warehouse, in that quantity. If the warehouse does not have that much free, the reservation is refused at this point rather than accepted and quietly broken later.

That refusal is the feature. A system that accepts every reservation and sorts it out at picking time has moved the problem to the worst possible moment, which is when someone is standing in the aisle.

## See what is held

“What stock is currently reserved?”

“How many blue chairs do we actually have free at Main?”

The first question is worth asking before promising a date to a customer, and the second during any investigation that starts with “we should have had enough of these.”

The second is answered from projected stock, where the reserved quantity reads your persisted active reservations. Where none exist, it falls back to open sales-order lines, so the projection is still meaningful before anyone has reserved anything explicitly.

## Release a hold

“Release the hold on those chairs, the order was cancelled.”

Releasing returns the quantity to available immediately. Cancelled orders, quotes that went cold, and holds placed against the wrong warehouse all end here.

Reservations are also released for you in the ordinary course of fulfilment: completing a pick list consumes the reservations it created, and cancelling one releases them. See putaway and pick lists for that flow, and item substitutes for what to reach for when the held item is short.

## Common questions

Does a reservation move the stock?
No. The quantity stays where it is and stays in your on-hand figures. What changes is how much of it anyone else is allowed to consume.

What happens if I try to reserve more than is available?
The reservation is refused, and the check uses actual quantity minus active reservations rather than raw on-hand. You find out at the moment you make the promise, not at the moment you break it.

Can I still issue stock that is reserved?
Not while the reservation is active. A `material_issue` that would breach it is blocked. Release the reservation first if the situation has genuinely changed.

Do reservations expire?
They stay until released or consumed. That is deliberate, because a hold that quietly lapses is worse than no hold at all, but it does mean stale reservations need clearing. `list-reservations` is the review tool.

How does this relate to pick lists?
Submitting a pick list places hard reservations for the quantities on it, completing the pick consumes them and produces a delivery note, and cancelling releases them. The two features are the same mechanism seen from different ends.

Where do I see the effect on planning?
`get-projected-qty` shows reserved quantity alongside actual, which is the figure to use when deciding whether to reorder.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-reservation --voucher-type sales_order --item CHAIR-BLUE --warehouse "Main" --qty 40
list-reservations
get-projected-qty --item CHAIR-BLUE --warehouse "Main"
release-reservation --id <reservation-id>`

See Core features

---

## Subcontract production to an outside manufacturer
URL: https://www.erpclaw.ai/docs/core/subcontracting/

# Subcontract production to an outside manufacturer
Send raw materials out, receive finished goods back, and let the books roll material and subcontracting costs into the finished item automatically.
Plenty of businesses make things without owning every machine. You buy the fabric and send it to a stitching unit, or ship components to a coating shop and get finished parts back. The accounting behind that is fiddly: your materials sit in someone else’s building but are still yours, and the finished item’s cost has to include both what you sent and what you paid for the work. ERPClaw runs the whole loop and keeps the books balanced at every step.

Say it the way you would out loud:

“We are sending 200 steel brackets to Apex Coating to be powder coated. Set that up.”

## Step 1, raise the subcontracting order

“Create a subcontracting order for Apex Coating: they turn our raw brackets into coated brackets.”

The order names the supplier, the finished item you expect back, and the quantity. Behind it sits the finished item’s recipe, which lists both the materials you will send and the coating service you are paying for. If the recipe does not list the service, the order is refused at submission rather than discovered broken at receiving time.

## Step 2, send the materials

“Ship the raw brackets for that order to Apex.”

The materials move out of your warehouse and into a supplier location that is still on your books, because the goods are still yours. Nothing is expensed yet. You can send in parts across several shipments if the truck fills up, and the order keeps track of what is still owed.

## Step 3, receive the finished goods

“Apex sent back 200 coated brackets. Their charge is 3 dollars each. Receive them.”

This is where the system earns its keep. One receipt does three things at once: the finished goods arrive into stock, their cost is set to the raw material you sent plus the subcontracting charge, and a draft bill for the charge is created for you to approve and pay through the normal purchasing flow. The materials at the supplier are consumed, the finished item carries the full honest cost, and the books balance without anyone doing journal-entry arithmetic.

Partial deliveries work the same way. The order moves through its life on its own: submitted, partially received, completed. The draft bill flows into the normal purchasing approval loop, and if the raw material itself runs short, item substitutes already know the ranked alternative.

## Changing your mind

Plans change before goods do. A subcontracting order can be cancelled outright while nothing has shipped. A materials shipment can be reversed as long as nothing has been received against it. Once finished goods have arrived, the posted receipt is permanent, and a correction is a reversal rather than an edit, which is how an honest ledger behaves.

## Common questions

Whose stock is it while it sits at the supplier?
Yours. The materials move to a supplier location that stays inside your books, so your stock reports still count them and nothing is written off just because it left the building.

How is the finished item’s cost worked out?
Raw material cost plus the subcontracting charge for the quantity received. The receipt posts once, with both parts included, so margin reports on the finished item are real from day one.

What if the supplier delivers in batches?
Receive each batch as it lands. The order tracks outstanding quantities for both the materials you send and the goods you get back, and moves to completed only when everything has arrived.

Do I pay the supplier through this flow?
The receipt creates a draft bill for the subcontracting charge. You approve and pay it exactly like any other supplier bill, so payment approvals stay in one place.

Can I cancel after sending materials?
You can reverse the material shipment while nothing has been received. After a receipt exists, cancellation means posting reversals, so the history stays intact.

What stops me from receiving these goods through a normal purchase receipt by mistake?
ERPClaw notices the link to the subcontracting order and routes the receipt through the subcontracting flow, so the finished goods are posted exactly once and never double counted.

## Reference: the actions behind this page

You never have to type these. They are what ERPClaw runs when you ask, and they are here for anyone calling the API or MCP surface directly, or reading a log.

`add-subcontracting-order
submit-subcontracting-order
transfer-materials-to-subcontractor
receive-subcontracted-items --subcontract-charge-rate <rate>
get-subcontracting-order / list-subcontracting-orders
cancel-subcontracting-order
cancel-subcontract-transfer`
The order lifecycle is draft, submitted, partially received, completed. A purchase receipt linked to a subcontracting order defers to `receive-subcontracted-items` so finished goods post exactly once.

See Core features

---

## Shopify App Store vs Custom App: Which Install Path?
URL: https://www.erpclaw.ai/docs/shopify/app-store-vs-custom-app/

# Shopify App Store vs Custom App: Which Install Path?
Comparison of the two ERPClaw Shopify install paths: Shopify App Store OAuth (recommended) vs Custom App with shpat_ token (power-user).
ERPClaw supports two ways to connect a Shopify store. Both lead to
the same outcome: orders, refunds, payouts, and disputes flowing into
your ERPClaw GL. The difference is plumbing.

## TL;DR

|
| App Store OAuth
| Custom App

| Setup
| 2 clicks + 1 CLI command
| Manual token creation in Partners dashboard

| Token rotation
| Handled automatically
| You manage it

| Worker dependency
| Yes (`shopify.erpclaw.ai`)
| No (skips Worker entirely)

| Background daemon
| Yes (status push every 15 min)
| No

| Status mirror in Shopify Admin
| Yes
| No

| Admin UI commands (Sync now, Disconnect)
| Yes
| No

| Best for
| Most merchants
| Air-gapped installs, power users, paranoid security teams

For 95% of merchants, App Store OAuth is the right choice. The
remaining 5% who care about the difference already know who they are.

## App Store OAuth (recommended)

This is the path you take when you click Add app on the ERPClaw
listing in the Shopify App Store.

Pros:

- Two-click install from a familiar interface

- OAuth token managed for you (no manual rotation)

- Status mirror in Shopify Admin so you see sync health at a glance

- Commands queued from Shopify Admin (Sync now, Disconnect) reach
your ERPClaw within seconds (SSE) or 15 minutes (status push)

- GDPR webhooks handled correctly

- Automatic reinstall handling on uninstall + reinstall within 48h

Cons:

- Requires our Worker to be reachable (`shopify.erpclaw.ai`). If our
Worker is down, your existing sync still works (the OAuth token is
on your machine), but new pairings and admin UI commands will fail.

- Adds a hop (Worker -> your ERPClaw) for command delivery, but not
for data sync.

See the install walkthrough for
the full step-by-step.

## Custom App (power-user)

This path skips the App Store and the Worker entirely. You create a
Shopify Custom App in your Partners dashboard, grant the same scopes
manually, and use the resulting `shpat_` access token directly.

Pros:

- Zero dependency on our Worker. ERPClaw talks straight to Shopify.

- Works in air-gapped environments (no inbound calls from our infra)

- You control the token rotation policy

- Useful when our Worker is down for maintenance and you cannot wait

- Clean fit for compliance environments where third-party
intermediaries are forbidden

Cons:

- More setup work (manual scope configuration, token generation)

- No status mirror in Shopify Admin (the admin UI shows “Custom App
install detected; status not available”)

- No daemon. You manage when sync runs (cron, scheduled tasks, manual)

- No remote command delivery. You run sync locally with `erpclaw shopify-sync-orders`

- You handle the GDPR webhook receiving infrastructure yourself if you
want to publish the app

### Custom App setup

-
In the Shopify Partners dashboard, create a new Custom App for
your store (Apps -> Create app -> Custom).

-
Configure the Admin API scopes: `read_orders`, `read_products`,
`read_customers`, `read_shopify_payments_accounts`.

-
Install the app to your store from within the dashboard.

-
Copy the resulting `shpat_xxxxxxxxxxxxxxxx` access token.

-
On your ERPClaw machine:

`erpclaw shopify-add-account \
--company-id 1 \
--shop-domain my-store.myshopify.com \
--access-token shpat_xxxxxxxxxxxxxxxx`

-
Verify the connection:

`erpclaw shopify-test-connection`

-
Run the first sync:

`erpclaw shopify-start-full-sync`

From this point, the workflow is identical to the OAuth flow except
that nothing pings our Worker.

## Mixing the two

You can run both flows on the same ERPClaw instance, even for the
same shop. The shopify_account table treats them as distinct accounts
identified by domain. We do not recommend mixing them for the same
shop because you will end up with duplicate sync activity, but it is
technically supported for testing.

## Switching between them

If you start with one flow and want to switch:

OAuth -> Custom: disconnect via the Shopify Admin Disconnect
button, uninstall the app from Shopify, then create a Custom App and
add via `shopify-add-account`. Your local data stays intact.

Custom -> OAuth: install the app from the Shopify App Store,
generate a pairing code, and run `shopify-connect`. The OAuth account
will be created alongside the existing custom account. Delete the old
custom account with `shopify-delete-account` once you confirm the
OAuth flow is working.

## When to choose which

Pick App Store OAuth if:

- You are a Shopify merchant who wants accounting automation and
doesn’t want to think about token management

- You want to see sync health from inside Shopify Admin

- You want push-button “Sync now” from Shopify Admin

Pick Custom App if:

- You run an enterprise IT environment that does not permit third-party
Worker dependencies

- You operate the app in an air-gapped network

- You want to script the entire sync lifecycle (cron, CI, custom orchestrator)

- You are auditing the integration for security and want to verify it
works without our infrastructure in the path

For everyone else: App Store OAuth.

See Shopify features

---

## Shopify Integration Architecture in ERPClaw
URL: https://www.erpclaw.ai/docs/shopify/architecture/

# Shopify Integration Architecture in ERPClaw
Technical architecture of the ERPClaw Shopify connector: Cloudflare Pages plus Worker plus your ERPClaw. Merchant data never transits our servers.
This page explains what runs where, which boundary each piece of data
crosses, and why the design puts merchant business data on the
merchant’s own machine instead of on our infrastructure.

## Three components

The connector has three moving parts and nothing else.

- Cloudflare Pages at `admin.shopify.erpclaw.ai`. Serves the
embedded admin UI (Polaris plus App Bridge v4). Loaded inside an
iframe by Shopify Admin.

- Cloudflare Worker at `shopify.erpclaw.ai`. A small Hono app
that handles the OAuth handshake, hands out pairing codes, receives
status pushes, verifies GDPR webhooks, and delivers queued commands.

- Your ERPClaw. Runs on a machine you control. Owns the merchant
business data: orders, customers, products, GL entries, inventory.
Talks to the Shopify Admin API directly using an OAuth token stored
locally.

` Shopify Admin (iframe)
|
v
Cloudflare Pages <----UI assets----
admin.shopify.erpclaw.ai
|
| App Bridge JWT
v
Cloudflare Worker <---OAuth, webhooks, status, commands---
shopify.erpclaw.ai
|
| pairing code redemption (one time)
| status push (every 15 min)
v
Your ERPClaw <===Shopify Admin API 2026-04===> Shopify
(SQLite or PostgreSQL on your machine) direct connection`
The critical invariant: merchant business data (orders, customers,
products) never transits the Worker. It flows on the bottom arrow,
directly between your ERPClaw and Shopify.

## Data flow during install (OAuth)

- Merchant clicks Add app in the Shopify App Store.

- Shopify redirects to `shopify.erpclaw.ai/oauth/install?shop=<shop>.myshopify.com`. The Worker validates the shop domain against a strict regex, sets a `__Host-` prefixed state cookie, and redirects to Shopify’s OAuth consent screen.

- Merchant approves. Shopify posts back to `/oauth/callback` with an authorization code and the state.

- The Worker verifies the state HMAC, exchanges the code for an offline access token, and stores it in Cloudflare KV under `pair:{code}` with a ten-minute TTL. Alongside the token it stores a per-shop HMAC secret derived via `HKDF-SHA256(master, salt=shop, info="erpclaw-status-v1", L=32)`.

- The admin UI loads and shows the six-character pairing code.

- Merchant runs `erpclaw shopify-connect --pairing-code <code>` locally. ERPClaw calls `GET /pair/<code>`. The Worker returns the OAuth token and the per-shop HMAC secret, then synchronously deletes the KV entry. The code is single-use.

- From this point ERPClaw holds the token. It calls `api.shopify.com/admin/api/2026-04/graphql.json` directly.

If the ten-minute window lapses, the code is gone and the merchant
generates a fresh one.

## Data flow during steady state

Your ERPClaw pulls Shopify data on its own schedule using its local
OAuth token. The Worker is out of the data path.

Every 15 minutes your ERPClaw POSTs a small status blob to
`shopify.erpclaw.ai/status/<shop>`. The blob contains:

- ERPClaw version string

- Timestamp of last successful sync

- Count of orders synced in the last 24 hours

- Count of GL entries posted in the last 24 hours

- Integer error count

- Local URL for the Open ERPClaw button

The push is signed with the per-shop HMAC and a timestamp (±300 second
tolerance). The Worker verifies it, writes it to KV under
`status:{shop}`, and uses it to render the admin UI status card. No
customer content, no order content, no financial amounts.

## Three latency tiers for command delivery

The admin UI can queue commands (Sync now, Disconnect). The
Worker delivers them to your ERPClaw in one of three ways:

Active tier (SSE): if your ERPClaw is online and holds an open
EventSource to `/events/<shop>`, the Worker pushes commands
immediately. Latency: a few hundred milliseconds.

Scheduled tier (15-minute push): when ERPClaw posts its next
status blob, the Worker piggybacks any queued commands in the response
body. Latency: up to 15 minutes. This is the default.

On-demand tier: the CLI command `erpclaw shopify-flush-pending-events`
forces an immediate round trip. Useful from scripts or cron jobs.

All three tiers use the same command schema and the same
acknowledgement mechanism. Delivery is at-least-once; commands carry
an `ack_id` for deduplication.

## GDPR webhooks

Shopify requires four compliance webhooks. All four hit the Worker.
The Worker verifies the `X-Shopify-Hmac-Sha256` signature
(constant-time compare against `SHOPIFY_CLIENT_SECRET`) before any
processing.

- `customers/data_request`: Worker holds no customer data; queues a `gdpr-dispatch` command for your ERPClaw.

- `customers/redact`: same path; your ERPClaw handles the actual redaction.

- `shop/redact`: fired 48 hours after uninstall. Worker deletes all `*:{shop}` KV keys. ERPClaw preserves GL entries (immutable; covered by GDPR Article 17(3)(b) exception for tax and accounting).

- `app/uninstalled`: Worker drops KV state immediately.

## Data location summary

| Data
| Where it lives
| Retention

| Shop domain, shop owner email
| Cloudflare KV (`meta:{shop}`)
| Until uninstall + 30 days

| OAuth access token
| Cloudflare KV (`pair:{code}`) during handoff only
| 10 min max, deleted on redemption

| Per-shop HMAC secret
| KV during handoff, then held by your ERPClaw
| 10 min on Worker, indefinite on your ERPClaw

| Status blob
| Cloudflare KV (`status:{shop}`)
| 30-day rolling

| Orders, customers, products
| Your ERPClaw database (SQLite default or PostgreSQL)
| Indefinite; you control it

| GL entries
| Your ERPClaw database (SQLite default or PostgreSQL)
| Indefinite; immutable

| Worker access logs
| Cloudflare Logpush to R2
| 90 days

## Source of truth

- Shopify connector code: github.com/avansaber/erpclaw-addons

- Privacy policy: avansaber.com/privacy-shopify

- Shopify Admin API version pinned: `2026-04`

See Shopify features

---

## Shopify GDPR Webhooks: Data Requests and Redaction
URL: https://www.erpclaw.ai/docs/shopify/gdpr-webhooks/

# Shopify GDPR Webhooks: Data Requests and Redaction
How the ERPClaw Shopify integration handles the four mandatory GDPR webhooks: data request, customer redact, shop redact, and app uninstalled.
Shopify mandates four compliance webhooks on every published app.
This page documents how ERPClaw handles each one, what data we hold
where, and what your obligations are as the data controller.

## The four webhooks

| Topic
| Fired when
| Worker behavior
| ERPClaw behavior

| `customers/data_request`
| Merchant requests data export for a customer
| Verifies HMAC, queues `gdpr-dispatch` for your ERPClaw, returns 200
| Generates JSON export of all customer-related rows, mails to merchant

| `customers/redact`
| 10 days after merchant-requested customer deletion
| Verifies HMAC, queues `gdpr-dispatch`, returns 200
| Redacts customer PII from local DB; preserves anonymized order/financial records

| `shop/redact`
| 48 hours after shop uninstall
| Verifies HMAC, deletes all `*:{shop}` KV keys, queues final `gdpr-dispatch` for your ERPClaw, returns 200
| Receives final cleanup signal; merchant decides what to redact locally

| `app/uninstalled`
| Immediately on app uninstall
| Verifies HMAC, drops KV state for the shop, returns 200
| Local ERPClaw data preserved; merchant retains control

Verification is constant-time HMAC-SHA256 against the
`SHOPIFY_CLIENT_SECRET` environment variable, per Shopify’s
webhook signature spec.

## Why the Worker forwards instead of acting

The Worker holds no merchant business data: no customer names, no
order amounts, no financial totals. So when Shopify asks “give me
everything you have about customer X”, the Worker has nothing to give.

But your ERPClaw does. So the Worker queues a `gdpr-dispatch` command
in Cloudflare KV with the shop, the topic, and the inbound payload.
Your ERPClaw picks it up on the next status check-in (or via SSE if
active), runs the actual data request or redaction against its local
SQLite database, and emails the result to the merchant.

This means:

- No customer data lives on our infrastructure, so we have no GDPR
Article 30 record-of-processing obligation for the bulk of the data.

- Your ERPClaw is the data controller, and you handle the GDPR
obligations for the data you hold.

- The Worker is only a relay; we hold data minimization to the
furthest extent possible.

## What ERPClaw redacts on `customers/redact`

When the `customers/redact` command reaches your ERPClaw, the
following happens locally:

- The customer’s `customer.name`, `customer.email`, `customer.phone`,
`customer.billing_address`, `customer.shipping_address` columns are
set to anonymized placeholders (e.g., `name = 'REDACTED-{customer_id}'`).

- Linked rows that contain customer PII (sales invoices, sales orders,
delivery notes) have their `customer_name` field anonymized too.

- GL entries are preserved. ERPClaw’s general ledger is immutable
by design; redacting GL entries would break the chain of custody and
violate accounting record-keeping requirements (US GAAP, IRS Pub 583,
GDPR Article 17(3)(b) exception for legal obligations).

- Stock ledger entries (SLE) are preserved for the same reason.

- The redaction event is logged in `data_audit_log` so you can prove
the request was honored.

## What `shop/redact` actually deletes

When `shop/redact` fires (48 hours after uninstall), the Worker
deletes:

- `meta:{shop}` (shop domain, owner email)

- `status:{shop}` (status blob with sync counters)

- `pair:*` keys still associated with the shop (should be empty by then)

- Any pending `cmd:*:{shop}` command queues

- Any pending `dedup:*:{shop}` deduplication state

ERPClaw receives a `gdpr-dispatch` command with topic `shop/redact`.
By design, ERPClaw does not delete your local SQLite data on this
signal. The merchant’s local install is theirs to control. You can
delete the local DB manually if you want a clean wipe.

## Data we never hold (GDPR negative-confirmation)

For audit purposes, here is what does not exist anywhere on our
Worker, never has, and never will:

- Customer names, emails, phone numbers, addresses

- Order line items, amounts, products

- Product details, SKUs, prices

- Inventory levels, warehouses

- Financial entries (invoices, payments, GL entries)

- Payout amounts (we hold counters, not amounts)

- Tax IDs, tax registration numbers

You can verify this by reading the Worker source at
github.com/avansaber/erpclaw-addons
and grepping for any of the above field names. They do not appear.

## Audit trail

If you need a GDPR Article 30 records-of-processing trail for an
audit, the following are available:

- Worker access logs: Cloudflare Logpush to R2, 90-day retention.

- GDPR webhook receipt log: every received webhook is logged with topic, shop, timestamp, and HMAC verification result. Available in `gdpr_webhook_log` Cloudflare KV index.

- ERPClaw `data_audit_log`: every redaction or data request your ERPClaw processes is logged locally with timestamp, customer_id, requestor email, and outcome.

## Where to test the flow

The `shopify shopify-handle-gdpr` action accepts a `--topic` and
`--payload` for offline testing. See the action documentation in the
SKILL.md source.

## Privacy policy

Our public privacy policy for the Shopify integration is at
avansaber.com/privacy-shopify.

For questions about the GDPR webhook flow, email
`support@avansaber.com`.

See Shopify features

---

## Shopify Integration Install Walkthrough
URL: https://www.erpclaw.ai/docs/shopify/install-walkthrough/

# Shopify Integration Install Walkthrough
Step-by-step Shopify accounting integration install for ERPClaw. From fresh Shopify store to first synced order in under 10 minutes.
This guide takes you from a fresh Shopify store to your first synced
order. Most merchants finish in under ten minutes.

## Prerequisites

You need three things: a Shopify store where you are the owner (or
have app install permission), a computer or small server where
ERPClaw will run (a laptop, a Mac mini, a cheap Linux VPS; anything
that stays online when you want to sync), and basic terminal access
on that machine. If “open a terminal and paste a command” is not
foreign to you, you are ready.

## 1. Install ERPClaw on your machine

ERPClaw runs on your own hardware. That is the whole point. Install
it with a single command:

`clawhub install erpclaw`
If you do not have `clawhub` yet, follow the short install guide at
docs.openclaw.ai/install. It
takes about two minutes.

Verify the install completed:

`python3 ~/.openclaw/workspace/skills/erpclaw/scripts/db_query.py --action status`
You should see a JSON status block with a non-zero `tables` count and
an `ok` status.

## 2. Install the Shopify app

Open the Shopify App Store listing for ERPClaw Accounting & ERP
and click Add app. Shopify shows you an OAuth consent screen
listing the permissions the app requests:

- `read_orders` so ERPClaw can pull your orders

- `read_products` so SKUs can map to your inventory

- `read_customers` so customer records can sync

- `read_shopify_payments_accounts` so payouts and disputes reconcile to your bank

All scopes are read-only. The app cannot change anything in your
store.

Approve the install. Shopify redirects you to the ERPClaw admin UI
embedded in your Shopify admin.

## 3. Get your pairing code

The first screen shows a six-character pairing code (for example,
`ABC-X9Z`) and a short instruction block.

The code is valid for ten minutes. If you miss that window, refresh
the page to generate a new one. Each code can only be used once.

## 4. Pair from the terminal

Copy the code. On the machine where ERPClaw is running, run:

`erpclaw shopify-connect --pairing-code ABC-X9Z`
Replace `ABC-X9Z` with your actual code. ERPClaw contacts our
Cloudflare Worker at `shopify.erpclaw.ai`, redeems the code, pulls
down the OAuth token, and stores it locally. You will see a success
message naming your shop domain.

From this moment, your ERPClaw talks to Shopify directly. Our servers
are not in the data path.

## 5. Run the first sync

Kick off a full initial sync:

`erpclaw shopify-start-full-sync`
ERPClaw pulls products, customers, orders, refunds, and payouts in
order. A small store finishes in under a minute. A store with tens
of thousands of orders can take a few minutes. The sync runs in the
background; you can close the terminal.

## 6. Use the dashboard

Go back to the Shopify admin tab. The status card shows the last 24
hours of activity:

- Orders synced. How many Shopify orders ERPClaw has pulled.

- GL entries posted. How many journal entries ERPClaw wrote to its double-entry general ledger.

- Errors. Problems during sync or posting. Zero is the healthy state.

- Last sync. When ERPClaw last reported in. Fresh is good; stale means your instance may be offline.

## 7. The three buttons

Open ERPClaw: deep-links to your local ERPClaw web UI so you can
browse the synced data.

Sync now: queues an on-demand sync. ERPClaw picks it up on its
next check-in (usually within seconds) and runs a fresh pull.

Disconnect: revokes the OAuth token, clears the pairing state,
and stops the push daemon. Your local ERPClaw data stays intact; only
the Shopify link is severed.

## 8. Where your data lives

Your orders, customers, products, and general ledger live in a SQLite
database at `~/.openclaw/erpclaw/data.sqlite` on your own machine.
Back it up like you would any other file. AvanSaber never sees this
data. Our Cloudflare Worker only holds your shop domain, a short-lived
pairing code, and a small status blob with sync counters.

---

## Shopify OAuth Pairing - How the App Store Flow Works
URL: https://www.erpclaw.ai/docs/shopify/oauth-pairing/

# Shopify OAuth Pairing - How the App Store Flow Works
The ERPClaw Shopify OAuth pairing flow: six-character codes, single-use redemption, per-shop HMAC, and App Bridge session tokens. Tokens never persist.
This page is for merchants and reviewers who want to understand
exactly how the Shopify App Store install flow works, what gets
stored where, and why the design eliminates a common class of
security incidents.

## The five-second version

When you click Add app in the Shopify App Store, three things
happen in order:

- Shopify and our Worker complete an OAuth handshake (offline access
token issued, HMAC verified, scopes accepted).

- Our Worker mints a six-character pairing code and shows it to you
in the embedded admin UI.

- You run `erpclaw shopify-connect --pairing-code ABC-X9Z` on your
own machine. The code is redeemed, the OAuth token is forwarded to
your ERPClaw, and the Worker deletes the token within 60 seconds.

After that, your ERPClaw talks to Shopify directly. The Worker is no
longer in the data path.

## Why pairing codes instead of polling

The most common pattern for OAuth + self-hosted clients is “tell the
client to come pick up the token via OAuth callback URL.” That breaks
when the client is behind NAT, has no inbound HTTPS, or is a CLI tool
without a web server.

ERPClaw uses out-of-band pairing: the user copies a short code
from one screen and pastes it into another. Same UX as Discord bot
linking, GitHub CLI device flow, or Apple TV sign-in. Works regardless
of the client’s network topology.

The code:

- Six characters from the unambiguous alphabet `ABCDEFGHJKLMNPQRSTUVWXYZ23456789` (no `0`/`O`/`1`/`I`)

- Single use

- Ten-minute TTL (Cloudflare KV expiration)

- Cryptographically random (32 bits of entropy; sufficient for a 10-min single-use exchange)

If a code is leaked or guessed, the worst case is the attacker pairs
to their own ERPClaw instead of yours. The token still gets deleted
from our Worker. Your store is unaffected.

## What lives on our Worker (and for how long)

| Item
| Location
| Lifetime

| Six-character pairing code
| Cloudflare KV `pair:{code}`
| 10 minutes max

| OAuth offline access token
| Cloudflare KV `pair:{code}`
| Deleted on first redemption (typically <60s)

| Per-shop HMAC secret
| Cloudflare KV `pair:{code}`
| Deleted with token; merchant’s ERPClaw retains it

| Shop domain
| Cloudflare KV `meta:{shop}`
| Until uninstall + 30 days

| Owner email
| Cloudflare KV `meta:{shop}`
| Until uninstall + 30 days

| Status blob
| Cloudflare KV `status:{shop}`
| 30-day rolling

Notably not on our Worker:

- Customer data (names, emails, addresses)

- Order data (amounts, line items, products)

- Product data (SKUs, prices, inventory)

- General ledger entries

- Any financial total in any form

Those live on your ERPClaw, full stop.

## App Bridge v4 session tokens

The embedded admin UI talks to our Worker using App Bridge v4 session
tokens (`getSessionToken()`). Each token is a JWT signed by Shopify
with the shop’s session secret. Our Worker:

- Verifies the JWT signature against the Shopify session-secret API

- Validates the `aud`, `iss`, and `exp` claims

- Caches verified tokens for up to 60 seconds (within the JWT’s exp window)

- Rate-limits per-shop requests

If you are integrating against our Worker directly (advanced use), see
the JWT verification source at
github.com/avansaber/erpclaw-addons.

## Per-shop HMAC secret derivation

Every shop has a unique HMAC secret used to sign status pushes from
your ERPClaw to our Worker. The secret is derived once via HKDF:

`shop_secret = HKDF-SHA256(
ikm: STATUS_HMAC_MASTER_KEY, // 32-byte secret, only on Worker
salt: shop, // e.g. "my-store.myshopify.com"
info: "erpclaw-status-v1",
length: 32 bytes
)`
The master key never leaves our Worker. Each shop’s derived secret is
known only to our Worker and to the merchant’s ERPClaw (delivered
once during pairing).

If we ever need to rotate the master key, every existing shop’s
derived secret rotates with it. Your ERPClaw will start getting 401s
on status pushes until you re-pair. We do not plan to rotate without
notice.

## What happens on uninstall

Two webhooks fire when you uninstall the app:

- `app/uninstalled` (immediate). Worker drops `meta:{shop}`,
`status:{shop}`, and any pending command queues. The next status
push from your ERPClaw will be rejected because there is no shop
record to validate against.

- `shop/redact` (48 hours later). Worker double-checks all
`*:{shop}` keys are cleared.

Your local ERPClaw data is not deleted. GL entries, orders,
customers, and products remain in your SQLite database. If you want
to delete them, do that locally; we have no remote way to reach into
your machine.

## Custom-app flow (skipping the Worker)

If you would rather not use the App Store flow at all, ERPClaw
supports the Shopify Custom App pattern. See the
App Store vs Custom App page
for the comparison.

In short: create a Custom App in your Shopify Partners dashboard,
grab the `shpat_` access token, and run:

`erpclaw shopify-add-account \
--company-id 1 \
--shop-domain my-store.myshopify.com \
--access-token shpat_xxxxx`
No Worker, no pairing code, no OAuth. You manage the token rotation
yourself.

## Source

- OAuth route source: `oauth.ts` in the Shopify Worker

- Pairing route: `pair.ts`

- Per-shop HMAC derivation: `hkdf.ts`

- Full Worker source: github.com/avansaber/erpclaw-addons

See Shopify features

---

## Shopify Integration Troubleshooting
URL: https://www.erpclaw.ai/docs/shopify/troubleshooting/

# Shopify Integration Troubleshooting
Fix common ERPClaw Shopify integration issues. Pairing code expired, OAuth scope errors, sync failures, status card stuck, webhook 401s, and reinstall flow.
Pick the symptom that matches. Each fix is short; most problems are a
stale pairing code or an offline ERPClaw instance.

## “App Bridge did not initialize” inside Shopify admin

This almost always means the app was opened via a direct URL instead
of through the Shopify Admin. App Bridge only initializes when loaded
inside the Shopify iframe with the correct session context.

Open the app the supported way:

`https://admin.shopify.com/store/<your-shop>/apps/erpclaw-accounting-erp`
If that still fails, clear your browser cache for `admin.shopify.com`
and reload.

## “Pairing code missing or expired”

Pairing codes live for ten minutes and can only be used once. If you
waited too long or the code was already redeemed, generate a fresh
one.

Open this URL in a browser where you are logged in to the Shopify
admin:

`https://shopify.erpclaw.ai/oauth/install?shop=<your-shop>.myshopify.com`
The admin UI reloads with a new code. Run
`erpclaw shopify-connect --pairing-code NEW-CODE` within ten minutes.

## OAuth fails with “invalid scope”

You are running an older build of the app that predates the umbrella
`read_shopify_payments_accounts` scope. Reinstall to get the current
scope set.

- Uninstall from Settings > Apps and sales channels in Shopify Admin.

- Wait a few seconds.

- Reinstall from the App Store.

## `shopify-connect` says “pairing not found”

The code was already redeemed, or the ten-minute window lapsed.
Pairing codes are single-use by design. Generate a fresh code (above)
and retry.

## First sync brings in zero orders

Two common causes. First, the store genuinely has no orders yet. Open
your Shopify Admin Orders tab and confirm. Second, the `read_orders`
scope was not granted, which can happen after a botched reinstall.

Verify and re-run:

`erpclaw shopify-test-connection
erpclaw shopify-start-full-sync`
If `shopify-test-connection` reports a scope problem, reinstall the
app from the App Store.

## Webhooks return 401 in my monitoring

A 401 from our Worker means the HMAC signature did not verify. Real
Shopify deliveries always sign correctly, so 401s are usually:

- A synthetic probe (uptime monitor, `curl` test) hitting the webhook endpoint without a valid signature. Expected and safe.

- A Worker secret rotation that drifted out of sync with Shopify. Rare.

Genuine Shopify webhook traffic from your store should always return
200. Verify by triggering a test webhook from **Settings > Notifications

Webhooks**.

## Status card stays stuck on a stale timestamp

The status card reflects the last push from your ERPClaw. If the
timestamp stops advancing, your push daemon is not running.

On Linux with systemd:

`systemctl --user status erpclaw-shopify-push`
On macOS:

`launchctl list | grep erpclaw`
If the daemon is missing, reinstall it:

`erpclaw shopify-install-daemon`
You can also force an immediate push:

`erpclaw shopify-push-status`

## Uninstall and reinstall

Shopify retains the install record for 48 hours after uninstall
because the `shop/redact` webhook fires on that delay. During that
window you can reinstall any time. After 48 hours all traces of your
shop are gone from our KV and the install starts from zero.

To reinstall immediately, go to the App Store listing and click
Add app again. Run through the pairing flow as if it were fresh.

## Where to get help

- Email `support@avansaber.com` with your shop domain, the ERPClaw version (`erpclaw shopify-status`), and the error message.

- Open a GitHub issue at github.com/avansaber/erpclaw-addons/issues.

- Source code for the Shopify integration is public; you can read exactly what happens at each step.

See Shopify features

---

## Stripe Integration Architecture in ERPClaw
URL: https://www.erpclaw.ai/docs/stripe/architecture/

# Stripe Integration Architecture in ERPClaw
Technical architecture of the ERPClaw Stripe connector. Direct API access from your machine. No middleware. Optional webhook receiver.
This page explains what runs where, what data is held where, and why
the design eliminates the typical “middleware syncs Stripe to
QuickBooks” pattern.

## Two components

The Stripe integration is simpler than the Shopify one. There are
only two moving parts:

- Your ERPClaw. Holds the API key, talks to the Stripe API
directly, writes journal entries to local SQLite.

- Stripe API at `api.stripe.com`. Source of truth for charges,
refunds, payouts, customers, subscriptions.

` Your ERPClaw <===Stripe API v2026-04===> Stripe
(SQLite on your machine)`
Optional third component: your ERPClaw’s webhook receiver, if
you want real-time sync. Stripe POSTs events to your receiver, your
ERPClaw verifies the signature and posts journal entries within
seconds. No third-party middleware, no message queues.

If you do not want to expose your ERPClaw to the internet, skip the
webhook and rely on polling sync (default 5-minute interval).

## What ERPClaw stores

Seventeen dedicated tables:

| Table
| Records

| `stripe_account`
| One per Stripe account you connect

| `stripe_charge`
| Every charge, success or failure

| `stripe_refund`
| Every refund, partial or full

| `stripe_dispute`
| Every dispute, including won/lost outcomes

| `stripe_payout`
| Every payout to your bank

| `stripe_payout_transaction`
| Per-transaction breakdown of each payout

| `stripe_customer`
| Every customer object

| `stripe_subscription`
| Every subscription with full lifecycle

| `stripe_invoice`
| Subscription and one-off invoices

| `stripe_invoice_line_item`
| Per-line-item detail

| `stripe_application_fee`
| Connect application fees

| `stripe_transfer`
| Connect transfers to connected accounts

| `stripe_balance_transaction`
| The Stripe ledger view

| `stripe_webhook_event`
| Webhook deduplication + audit

| `stripe_sync_job`
| Sync run history with timing

| `stripe_gl_rule`
| Configurable routing rules per event type

| `stripe_reconciliation_run`
| Per-payout reconciliation runs

Plus 12 dedicated GL accounts auto-created on connect:

- `stripe_clearing`, funds Stripe holds before payout

- `stripe_revenue`, gross revenue

- `stripe_fee`, Stripe processing fees

- `stripe_refund`, refund holding (offset)

- `stripe_chargeback`, disputed charges

- `stripe_chargeback_fee`, Stripe’s chargeback fee

- `stripe_deferred_revenue`, ASC 606 liability

- `stripe_application_fee`, Connect platform revenue

- `stripe_connect_transfer`, Connect transfers held

- `stripe_reserve`, Stripe reserves (rare)

- `stripe_dispute_reserve`, held during dispute

- `stripe_bank`, your bank account from payouts

## Sync modes

Polling (default): a background daemon runs `stripe-sync-charges`,
`stripe-sync-refunds`, and `stripe-sync-payouts` every 5 minutes.
Catches everything within 5 minutes of when Stripe records it.

Webhook (real-time): register a webhook endpoint with
`stripe-register-webhook`. Stripe posts events to your ERPClaw within
seconds. Your ERPClaw verifies the signature with HMAC-SHA256 against
the webhook secret, deduplicates by event ID, and posts journal
entries.

Manual: trigger any sync command on demand from CLI or chat
interface.

You can run all three modes simultaneously. Webhook events take
precedence; polling fills any gaps; manual catches edge cases.

## Webhook signature verification

When ERPClaw receives a webhook, it:

- Extracts the `Stripe-Signature` header

- Splits into timestamp + signatures

- Verifies the timestamp is within 5 minutes of now (replay protection)

- Concatenates `timestamp.payload` and HMAC-SHA256 with the webhook secret

- Constant-time compares to the signature

- Looks up the event ID in `stripe_webhook_event`; if present, returns 200 (deduplication); if not, processes and records

Standard Stripe pattern; we do not deviate.

## API rate limits

Stripe rate-limits API calls per account: 100 read requests per
second in live mode, 25 in test mode. Bursts above that get 429
responses. ERPClaw uses exponential backoff with jitter, capped at
60 seconds, for any 429.

For initial syncs of large accounts (millions of charges), the bulk
import endpoints are used where available. A 10-million-charge sync
typically completes in 30-90 minutes depending on Stripe’s response
times.

## Connect platforms

If you run a marketplace on Stripe Connect, ERPClaw handles the three
standard patterns:

- Direct charges with `application_fee_amount`

- Destination charges with `transfer_data.destination`

- Separate charges and transfers for full control

Each pattern posts journal entries to the correct combination of
clearing, revenue, application_fee, and connect_transfer accounts.
See Connect platform fees.

## ASC 606 revenue recognition

If you sell subscriptions, ERPClaw can be configured to post charges
to Deferred Revenue (a liability) instead of immediate revenue,
then recognize revenue monthly over the subscription period. See
ASC 606 reference.

## Source

- Stripe connector code: github.com/avansaber/erpclaw-addons

- Stripe API version pinned: `2026-04`

See Stripe features

---

## Stripe ASC 606 Revenue Recognition for SaaS
URL: https://www.erpclaw.ai/docs/stripe/asc-606/

# Stripe ASC 606 Revenue Recognition for SaaS
How ERPClaw handles Stripe ASC 606 revenue recognition. Deferred revenue, monthly recognition, mid-cycle changes, refunds, and the journal entries that prove it works.
ASC 606 is the accounting standard for revenue recognition. It says
you cannot count an annual subscription as revenue when the customer
pays it. You have to recognize it monthly over the subscription
period.

This page documents how ERPClaw implements ASC 606 for Stripe
subscriptions, with the actual journal entries.

## The five steps

ASC 606 defines a five-step framework for revenue recognition. For
Stripe subscriptions, the steps map as follows:

- Identify the contract: a Stripe subscription = a contract.

- Identify performance obligations: the obligation is to deliver software access for the subscription period.

- Determine the transaction price: the subscription’s recurring amount.

- Allocate the transaction price: for single-product subscriptions, all of it goes to the one obligation.

- Recognize revenue when (or as) the obligation is satisfied: monthly, over the subscription period.

Multi-product subscriptions and bundles need step 4 to allocate, which
gets complicated. ERPClaw handles single-product subscriptions
correctly out of the box; multi-product allocations require manual
configuration of allocation ratios.

## The journal entries

### Customer pays $1,200 for an annual subscription

When the charge succeeds:

| Account
| Debit
| Credit

| Stripe Clearing
| $1,200
|

| Stripe Fees
| $35
|

| Deferred Revenue
|
| $1,200

| Stripe Fee (Expense)
|
| $35 (offset)

Note: the Stripe fee is an immediate expense, not deferred.

### End of month 1 (recognize 1/12)

| Account
| Debit
| Credit

| Deferred Revenue
| $100
|

| Subscription Revenue
|
| $100

Repeat each month.

### End of month 12 (final recognition)

| Account
| Debit
| Credit

| Deferred Revenue
| $100
|

| Subscription Revenue
|
| $100

After this, Deferred Revenue for this subscription is $0.

### Customer cancels mid-year (after 4 months)

Eight months unused at $100/month = $800 unrecognized.

If refunded:

| Account
| Debit
| Credit

| Deferred Revenue
| $800
|

| Stripe Clearing
|
| $800

If not refunded (e.g. cancellation effective end of period):

| Account
| Debit
| Credit

| Deferred Revenue
| $800
|

| Subscription Revenue
|
| $800

The full $800 is recognized at cancellation because the obligation is
discharged.

## Mid-cycle upgrades and downgrades

If a customer upgrades from $100/mo to $200/mo on day 15 of a 30-day
cycle, the proration creates a charge for the difference applied
prorated to the remaining 15 days.

ERPClaw posts the new charge to Deferred Revenue normally, then
adjusts the recognition schedule:

- Days 1-15: revenue recognized at $100/30 per day

- Days 16-30: revenue recognized at $200/30 per day

- Days 16-30 also recognize the proration top-up

The math is in `stripe_subscription_recognition_schedule` table; you
can audit per-subscription, per-day recognition by inspecting it.

## Annual subscriptions paid with discount

If a customer pays $1,000 for a $1,200 annual plan (Stripe Coupon
applied), the journal entries reflect the actual cash:

| Account
| Debit
| Credit

| Stripe Clearing
| $1,000
|

| Deferred Revenue
|
| $1,000

Monthly recognition is then $83.33, not $100. The discount is
absorbed into the recognized revenue, not separately accounted for.

## Failed payment then re-attempted successfully

Stripe attempts to charge, fails, retries, succeeds. ERPClaw posts
journal entries only on successful charge. The failed attempt
creates a `stripe_charge` row with `status='failed'` for audit
purposes but no GL impact.

## Subscriptions on Stripe trial

Trial periods do not collect payment. No journal entry is posted
during trial. When the trial ends and the first paid charge succeeds,
revenue recognition begins from that date.

If the customer cancels during trial, no journal entry, no Deferred
Revenue.

## Multi-currency subscriptions

ERPClaw v1 supports USD-denominated subscriptions only. Other
currencies are recorded but not auto-converted to USD for GL
posting. Multi-currency runs payment-side across 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED): an invoice settles in the currency it was raised in, with no conversion inside a transaction.

## Audit trail

For an auditor reviewing your books, ERPClaw provides:

- `stripe_subscription_recognition_schedule`: per-subscription, per-day
recognition planned and actual

- `gl_entry` joined to `stripe_subscription`: every recognition entry
is linked back to the subscription that triggered it

- `data_audit_log`: every change to a recognition schedule (proration,
cancellation, refund) is logged with timestamp and trigger event

You can export any subset for a CPA or auditor:

`erpclaw stripe-export-asc606 --account-id <id> --period-start 2026-01-01 --period-end 2026-04-30`
Outputs a CSV per accountant convention.

## Edge cases handled

- Proration on upgrade

- Proration on downgrade (results in credit, applied to next invoice)

- Refund of a partially recognized subscription

- Cancellation with end-of-period grace

- Payment failure followed by recovery

- Subscription resumed after pause

- Trial-to-paid transition

- Free plan to paid upgrade

## Edge cases NOT handled in v1

- Multi-currency conversions

- Custom billing cycles beyond month/year (weekly, daily, custom day count)

- Multi-element subscriptions with manual allocation

- Variable consideration adjustments (price changes mid-cycle by usage)

If you hit one of these, the recognition will need a manual journal
entry. Email `support@avansaber.com` if you have a regular case that
isn’t covered.

## ASC 606 in plain English

You are not paying for “access to a SaaS tool.” You are paying for
12 months of access. The vendor cannot count your money as
revenue until they have actually given you access for the period
they promised. ASC 606 makes this rule formal.

The accounting effect is straightforward but easy to get wrong by
hand. ERPClaw automates it for any size book.

## Source

- ASC 606 standard: fasb.org

- Stripe docs on revenue recognition: stripe.com/docs/billing/revenue-recognition

- ERPClaw recognition engine source: github.com/avansaber/erpclaw-addons

See Stripe features

---

## ERPClaw Stripe Connect Fees: Fields, Flows, and Events
URL: https://www.erpclaw.ai/docs/stripe/connect-platform-fees/

# ERPClaw Stripe Connect Fees: Fields, Flows, and Events
Configure ERPClaw for Stripe Connect direct charges, destination charges, and separate charges and transfers, including fields, events, refunds, transfers, and payouts.
If you run a marketplace or platform on Stripe Connect, you have an
extra layer of complexity in your accounting: application fees,
transfers, reverse transfers, and connected account payouts.

This page documents the three Stripe Connect patterns and how ERPClaw
handles each.

Use this reference for Stripe objects, ERPClaw configuration, events,
refund flags, and ledger flow. For the accounting analysis and worked
journal logic, read the Stripe Connect accounting guide.

Primary Stripe references: direct charges, destination charges, separate charges and transfers, the Connect charge types overview, and refund and dispute handling.

## The three Connect patterns

Stripe Connect supports three ways to split a payment between your
platform and a connected account.

### Pattern 1: Direct charges

The customer’s payment is charged on the connected account’s books.
Your platform takes a cut as `application_fee_amount`. Stripe
deducts its processing fee from the connected account in this worked
example, where that account is configured as the Stripe fee payer.

Customer pays $100. Your application fee is $10. Stripe fee is $3.20.

Connected account’s books:

| Account
| Debit
| Credit

| Stripe Clearing
| $86.80
|

| Stripe Fees
| $3.20
|

| Application Fee
| $10.00
|

| Revenue
|
| $100.00

Your platform’s books (separate ERPClaw account):

| Account
| Debit
| Credit

| Stripe Clearing
| $10.00
|

| Application Fee Revenue
|
| $10.00

Use `--pattern direct` when configuring Connect.

### Pattern 2: Destination charges

The customer’s payment is charged on your platform’s books. You
specify a `transfer_data.destination` (the connected account) and a
`transfer_data.amount` (how much to transfer).

Customer pays $100. You transfer $90 to the connected account
(retaining $10 as platform revenue). Stripe’s processing fee comes off
your platform’s side.

Your platform’s books:

| Account
| Debit
| Credit

| Stripe Clearing
| $96.80
|

| Stripe Fees
| $3.20
|

| Revenue
|
| $100.00

The transfer is a separate event:

| Account
| Debit
| Credit

| Connect Transfer
| $90.00
|

| Stripe Clearing
|
| $90.00

The transfer is a separate event that hits clearing and connect
transfer.

Connected account’s books:

| Account
| Debit
| Credit

| Stripe Clearing
| $90.00
|

| Revenue
|
| $90.00

Use `--pattern destination` when configuring Connect.

### Pattern 3: Separate charges and transfers

You charge on your platform first, then transfer separately at any
later time. Maximum control.

Customer pays $100, charged on your platform:

| Account
| Debit
| Credit

| Stripe Clearing
| $96.80
|

| Stripe Fees
| $3.20
|

| Revenue
|
| $100.00

Later, you transfer $90 to the connected account:

| Account
| Debit
| Credit

| Connect Transfer Pending
| $90.00
|

| Stripe Clearing
|
| $90.00

When the transfer settles on the connected side, ERPClaw posts on
the connected account:

| Account
| Debit
| Credit

| Stripe Clearing
| $90.00
|

| Revenue
|
| $90.00

And on the platform side, the pending transfer becomes an actual
transfer:

| Account
| Debit
| Credit

| Connect Transfer
| $90.00
|

| Connect Transfer Pending
|
| $90.00

Use `--pattern separate` when configuring Connect.

## Refunds in each pattern

### Direct charge refund

The refund happens on the connected account. Application fee may or
may not be refunded depending on `refund_application_fee=true`.

If application fee refunded:

Connected account:

| Account
| Debit
| Credit

| Stripe Clearing
| $10.00
|

| Application Fee
|
| $10.00

The customer refund is a separate event:

| Account
| Debit
| Credit

| Revenue
| $100.00
|

| Stripe Clearing
|
| $100.00

The original $3.20 processing fee remains in Stripe Fees. It is not
reversed under standard pricing.

Platform:

| Account
| Debit
| Credit

| Application Fee Revenue
| $10.00
|

| Stripe Clearing
|
| $10.00

If application fee NOT refunded (you keep your cut even on refund):

Platform: no journal entry. The $10 stays on your books.

### Destination charge refund

Refund happens on platform. Optionally `reverse_transfer=true` to
also reverse the transfer to the connected account.

If reverse_transfer=true:

Platform:

| Account
| Debit
| Credit

| Revenue
| $100.00
|

| Stripe Clearing
|
| $100.00

Then a Connect Transfer reversal moves the $90 back from the connected
account to platform clearing. The original $3.20 processing fee remains
in Stripe Fees, so the platform still bears that cost.

If reverse_transfer=false: connected account keeps the $90; platform
absorbs the loss.

### Separate refund

Reverse the charge first, then optionally reverse the transfer.
Most flexible but requires more discipline.

## Application fee refund only

Stripe lets you refund only the application fee without refunding the
charge itself. This is useful when you over-charged a marketplace fee.

Platform:

| Account
| Debit
| Credit

| Application Fee Revenue
| $5.00
|

| Stripe Clearing
|
| $5.00

The $5 is moved from your platform back to the connected account.

## Multi-account dashboard

If you connect multiple Stripe accounts (your platform plus several
connected accounts), each gets its own ERPClaw `stripe_account` row
with its own GL account namespace. Reports can be run per-account or
consolidated.

`erpclaw stripe-list-accounts
erpclaw stripe-revenue-summary --account-id <platform-id>
erpclaw stripe-revenue-summary --account-id <connected-id>`

## When to choose which pattern

Direct charges: when each connected account has its own customer
relationship, brand, and tax ID. Marketplace style (Etsy, Airbnb).

Destination charges: when your platform handles the customer
relationship, your tax ID is on the receipt, and the connected
account is just a payee. SaaS marketplace style (your platform sells,
splits revenue with vendor).

Separate charges and transfers: when you need full control over
when transfers happen (delayed payouts, escrow, hold-and-release
flows).

ERPClaw supports all three, but you choose one per Connect platform
during setup. Mixing patterns within one platform is technically
possible but produces messy books.

## Source

- Stripe Connect docs: stripe.com/docs/connect

- ERPClaw Connect handler: github.com/avansaber/erpclaw-addons

See Stripe features

---

## Stripe Integration Install Walkthrough
URL: https://www.erpclaw.ai/docs/stripe/install-walkthrough/

# Stripe Integration Install Walkthrough
Step-by-step Stripe accounting integration install for ERPClaw. From fresh Stripe account to first synced charge in under 5 minutes.
This guide takes you from a fresh Stripe account to your first synced
charge with proper journal entries. Most SaaS founders finish in
under five minutes.

## Prerequisites

You need three things: a Stripe account where you have at least
read-only access to the Dashboard, a computer or small server where
ERPClaw will run, and basic terminal access on that machine.

If you have not installed ERPClaw yet, see the
ERPClaw install guide first.

## 1. Get a Stripe restricted key (recommended)

We strongly recommend a restricted API key rather than a
secret key. ERPClaw needs read access to the following resources:

- Charges, refunds, disputes (read)

- Payouts (read)

- Customers (read)

- Subscriptions, invoices (read)

- Connect: Application fees, transfers (read; only if using Connect)

- Webhook endpoints (write; for the webhook setup)

In the Stripe Dashboard, go to Developers > API Keys > Create
restricted key. Name it `erpclaw-readonly`. Set the permission
matrix to read-only on the resources above. Save and copy the
`rk_live_xxxxxxxxxxxxxxxx` token.

For testing, use a `rk_test_xxxxx` key from your test mode.

## 2. Add the Stripe account to ERPClaw

`erpclaw stripe-add-account \
--company-id 1 \
--account-name "Acme SaaS" \
--api-key rk_live_xxxxxxxxxxxxxxxx`
ERPClaw verifies the key works by fetching your Stripe account info,
then creates the 12 GL accounts for Stripe (clearing, revenue, fees,
refund, chargeback, chargeback_fee, deferred_revenue, application_fee,
connect_transfer, reserve, bank, dispute_reserve).

## 3. Run the first sync

`erpclaw stripe-start-full-sync --account-id <id-from-step-2>`
ERPClaw pulls customers, charges, refunds, disputes, payouts, and
subscriptions in order. A small SaaS finishes in seconds. A larger
account with hundreds of thousands of transactions can take a few
minutes.

## 4. Verify GL entries

Check that journal entries are posting:

`erpclaw stripe-list-charges --account-id <id> --limit 5
erpclaw stripe-list-payouts --account-id <id>
erpclaw stripe-clearing-balance --account-id <id>`
The clearing balance should be very close to your in-flight Stripe
balance (charges that have not yet been paid out). If it matches
within rounding, your books are correct.

## 5. Set up webhook (recommended)

For real-time sync, register an endpoint:

`erpclaw stripe-register-webhook --account-id <id>`
This creates a Stripe webhook endpoint pointing at your ERPClaw’s
webhook receiver. Whenever Stripe fires an event (new charge, refund,
dispute, etc.), your ERPClaw posts the journal entries within
seconds.

If you cannot expose your ERPClaw to the public internet, skip this
step and rely on the polling sync (`stripe-start-full-sync` on a
cron).

## 6. Configure ASC 606 revenue recognition (SaaS only)

If you sell subscriptions, you almost certainly need ASC 606
treatment. Enable it:

`erpclaw stripe-configure-asc606 --account-id <id> --enabled`
ERPClaw will:

- Post subscription payments to Deferred Revenue instead of Revenue

- Recognize a portion of revenue every month over the subscription period

- Handle mid-cycle changes (upgrades, downgrades, refunds) correctly

See ASC 606 reference for the details.

## 7. Configure Connect (platforms only)

If you run a marketplace on Stripe Connect:

`erpclaw stripe-configure-connect --account-id <id> --pattern direct`
Patterns: `direct` (direct charges with application_fee), `destination`
(destination charges), or `separate` (separate charges and transfers).
See Connect platform fees for
the accounting treatment.

## Where your data lives

Your charges, customers, payouts, and general ledger live in a SQLite
database at `~/.openclaw/erpclaw/data.sqlite` on your own machine.
Back it up like any other file. AvanSaber never sees this data.

---

## Stripe MRR Reports - SaaS Metrics from Your Stripe Data
URL: https://www.erpclaw.ai/docs/stripe/mrr-reports/

# Stripe MRR Reports - SaaS Metrics from Your Stripe Data
How ERPClaw computes MRR, ARR, churn, expansion, and contraction from your Stripe subscription data. Cohort retention, net new MRR, and the math that powers SaaS reports.
If you run a SaaS business on Stripe, MRR is the metric that defines
your trajectory. ERPClaw computes MRR, ARR, churn, expansion, and
contraction directly from your Stripe subscription data, not from
estimates or aggregations.

## What ERPClaw reports

You ask for these the way you would ask a finance lead:

“What is our MRR this month, and how much of the change was expansion versus churn?”

The table below names the same metrics and the action behind each one, for when you are calling them directly.

| Metric
| Definition
| Command

| MRR
| Sum of all active subscription monthly amounts
| `stripe-mrr`

| ARR
| MRR × 12
| `stripe-arr`

| New MRR
| MRR added by brand-new subscriptions in the period
| `stripe-mrr-breakdown`

| Expansion MRR
| MRR added by upgrades + add-ons
| `stripe-mrr-breakdown`

| Contraction MRR
| MRR lost from downgrades
| `stripe-mrr-breakdown`

| Churn MRR
| MRR lost from cancellations
| `stripe-mrr-breakdown`

| Net New MRR
| New + Expansion − Contraction − Churn
| `stripe-mrr-breakdown`

| Logo churn
| Customers cancelled / customers at start of period
| `stripe-churn`

| Revenue churn
| Churn MRR / MRR at start of period
| `stripe-churn`

| Cohort retention
| Per-cohort retention curve over months
| `stripe-cohort-retention`

| Customer lifetime value
| Average revenue per customer × average lifespan
| `stripe-ltv`

All metrics are computed from `stripe_subscription` and
`stripe_subscription_item` tables, not from aggregate Stripe API
calls.

## How MRR is calculated

MRR is the sum of all currently-active subscriptions, normalized to
one month.

For each active subscription:

- If billing interval is monthly: add the unit_amount × quantity

- If billing interval is yearly: add (unit_amount × quantity) / 12

- If billing interval is weekly: add (unit_amount × quantity) × 52 / 12

- If billing interval is daily: add (unit_amount × quantity) × 365 / 12

Multi-line subscriptions (a base plan plus add-ons): each line is
summed separately.

Trialing subscriptions: NOT included in MRR until the first paid
charge succeeds. (Some companies count trials in MRR; we do not, and
we think the inclusion masks churn.)

Paused subscriptions: NOT included.

Past-due (delinquent): included for the first 7 days, then excluded.
You can override this in `stripe-configure-mrr-policy`.

## How churn is calculated

Two definitions, both reported:

Logo churn = (customers who cancelled in period) / (customers active at start of period). What percentage of your customers left.

Revenue churn = (sum of MRR lost to cancellation in period) / (MRR at start of period). What percentage of your revenue left.

Logo churn and revenue churn diverge when high-paying customers leave
disproportionately (or stay disproportionately). Both matter; either
can mask the other.

## Net new MRR formula

`Net New MRR = New MRR + Expansion MRR − Contraction MRR − Churn MRR`
Components:

- New MRR: subscriptions where the customer signed up in this
period AND the subscription started AND first charge succeeded

- Expansion MRR: existing customers whose MRR went up (upgrades,
add-ons, plan-tier increases)

- Contraction MRR: existing customers whose MRR went down (downgrades,
plan-tier decreases) but still subscribed

- Churn MRR: subscriptions that cancelled or expired in this
period

Period is configurable; default is calendar month.

## Cohort retention

ERPClaw groups customers by signup month (the cohort), then tracks
each cohort’s retained MRR over the following months.

`erpclaw stripe-cohort-retention --account-id <id> --periods 12`
Output is a 2D table:

| Cohort
| M0
| M1
| M2
| M3
| M4
| M5
| M6

| 2026-01
| $5,000
| $4,800
| $4,650
| $4,500
| $4,350
| $4,200
| $4,000

| 2026-02
| $7,000
| $6,700
| $6,500
| $6,300
| $6,150
| $5,950
| Not available

| 2026-03
| $9,500
| $9,100
| $8,900
| $8,700
| $8,500
| Not available
| Not available

Each row shows what that cohort’s MRR is N months after they signed
up. Decay over time = churn. Increase = expansion (rare; usually due
to upgrades).

## LTV calculation

`stripe-ltv` reports lifetime value as average revenue per customer
multiplied by average customer lifespan (in months).

`LTV = (sum of revenue per customer) / (count of customers) × (sum of months active per customer) / (count of customers)`
For customers who haven’t churned yet, lifespan is computed up to
today. As cohorts mature, the LTV becomes more accurate.

ERPClaw also reports LTV/CAC if you have configured
acquisition cost data via `stripe-set-cac` (per-month or per-channel).

## Annual subscriptions and MRR

Annual subscriptions normalized to MRR can mislead. A customer
paying $12,000/year is $1,000 MRR. If they don’t renew, that’s $1,000
MRR churn. But the cash impact is $12,000 over 12 months, not $1,000.

ERPClaw reports both:

- Booked MRR (the normalized figure used for SaaS metrics)

- Cash MRR (actual cash collected, divided by months in the period)

Use Booked MRR for investor reports. Use Cash MRR for runway models.

## Multi-currency

ERPClaw v1 reports MRR in USD. Subscriptions in other currencies are
recorded but not converted, deliberately: ERPClaw never converts currency inside a transaction, so each subscription’s MRR reads in the currency it bills in.

## Common edge cases

- Quarterly subscriptions: normalized to MRR by dividing by 3.

- Custom billing cycles: if not month/year/week/day, fall back to (unit_amount × quantity) × 12 / billing_period_in_months.

- Free plans: not in MRR by definition. They appear in “active customers” count but contribute $0.

- Discounts: applied to MRR. A $100 plan with 50% off is $50 MRR.

- Coupons that expire: ERPClaw recomputes MRR when the coupon
expires; the customer’s MRR jumps back to full price on that date.

- Subscription paused mid-cycle: subtracted from MRR on the pause
date.

## Audit trail

Every MRR computation is reproducible. Run with `--show-detail` to
get the per-customer breakdown:

`erpclaw stripe-mrr --account-id <id> --as-of 2026-04-30 --show-detail`
Output is a CSV with every active subscription, its MRR contribution,
and the math used.

## When MRR doesn’t match Stripe Dashboard

Stripe Dashboard’s MRR can differ from ERPClaw’s because:

- Stripe includes trials in some views; ERPClaw does not

- Stripe handles paused subscriptions differently

- Stripe rounds MRR per subscription; ERPClaw computes exactly

If they differ by more than 5%, run `stripe-mrr-breakdown` to see
which subscriptions are causing the gap.

## Source

- MRR engine source: github.com/avansaber/erpclaw-addons

- SaaS metrics primer: klipfolio.com/resources/articles/what-is-mrr

See Stripe features

---

## Stripe Integration Troubleshooting
URL: https://www.erpclaw.ai/docs/stripe/troubleshooting/

# Stripe Integration Troubleshooting
Fix common ERPClaw Stripe integration issues. API key errors, sync failures, webhook 401s, payout reconciliation gaps, ASC 606 imbalances, MRR misreports.
Pick the symptom that matches.

## “Stripe API authentication failed”

Either the API key is wrong, the key was rotated in the Stripe
Dashboard, or you are using a `pk_` key (publishable, client-side)
instead of `rk_` (restricted, server-side) or `sk_` (secret).

Verify by listing your accounts:

`erpclaw stripe-list-accounts`
If your account shows but sync fails, rotate the key in Stripe
Dashboard and update ERPClaw:

`erpclaw stripe-update-account --account-id <id> --api-key rk_live_NEW_KEY`

## Sync stalls or returns 429

Stripe rate-limits at 100 read requests per second in live mode. If
ERPClaw is mid-sync of a very large account and you also have other
processes hammering Stripe, you will see 429s. ERPClaw retries with
exponential backoff, so eventually completes. To speed up, pause
other processes.

If sync is permanently stuck, check the sync job:

`erpclaw stripe-list-sync-jobs --status running
erpclaw stripe-get-sync-job --sync-job-id <id>`
Cancel and retry:

`erpclaw stripe-cancel-sync-job --sync-job-id <id>
erpclaw stripe-start-full-sync --account-id <id>`

## Webhook returns 400 Bad Request

Stripe sends webhooks signed with HMAC. If your ERPClaw rejects them,
either the webhook secret is wrong (key rotated, accidentally pasted
the wrong one) or the request is being modified in transit (unusual
proxy, MITM testing).

Re-register the webhook to issue a fresh secret:

`erpclaw stripe-register-webhook --account-id <id>`
This deletes the old endpoint in Stripe and creates a new one with
a new signing secret stored locally.

## Webhook returns 200 but events are missing

Check the deduplication table:

`erpclaw stripe-list-webhook-events --account-id <id> --status processed --limit 20`
If events are processed but no journal entries appear, check the
GL routing rules for that event type:

`erpclaw stripe-list-gl-rules --account-id <id> --transaction-type charge`
If no rule exists for the event type, ERPClaw logs the event but does
not post to GL. Either add a rule or use the default rules
(re-add account triggers default-rule install).

## Payout reconciliation does not balance

Run the three-layer reconciliation report:

`erpclaw stripe-run-reconciliation --account-id <id> --period-start 2026-04-01 --period-end 2026-04-30`
The report shows Layer 1 (transactions to journal entries), Layer 2
(payouts to bank deposits), and Layer 3 (clearing account net).

Common causes of imbalance:

- A bank deposit was manually entered with a different amount than the actual Stripe payout

- A transaction was posted manually before sync caught it (causing a duplicate)

- A test-mode transaction got posted in live mode (or vice versa)

- A refund was posted but the original charge sync was missed

The reconciliation report shows which specific transactions are
unmatched. Fix those manually with `stripe-match-bank-transaction` or
by deleting the duplicate journal entry and resyncing.

## Clearing balance does not match Stripe Dashboard

Stripe’s “Available balance” minus “Pending balance” should equal
your Stripe Clearing GL account. If it doesn’t:

- Run `erpclaw stripe-clearing-balance --account-id <id>` and note the value

- Check the Stripe Dashboard’s Balance overview

- Difference is usually pending payouts not yet recorded in your bank

Force a fresh payout sync:

`erpclaw stripe-sync-payouts --account-id <id>`

## ASC 606 deferred revenue does not balance

If you have ASC 606 enabled and the Deferred Revenue account does not
match the sum of unrecognized subscription periods, run the
reconciliation:

`erpclaw stripe-asc606-reconcile --account-id <id> --period-end 2026-04-30`
Common causes:

- Mid-cycle subscription change not handled

- Refund of a partially-recognized subscription

- Multi-currency subscriptions (we support USD only in v1)

- Custom billing cycles outside the standard month/year patterns

The reconcile command shows which subscriptions are causing the
imbalance.

## MRR report shows wrong number

ERPClaw computes MRR from your subscription data, not from your
revenue. If MRR looks wrong:

- Check that paused subscriptions are excluded

- Check that trial subscriptions are excluded (until first charge)

- Check that proration is computed at the active rate, not the
initial rate

- Check that annual subscriptions are normalized to /12

Run the detailed MRR breakdown:

`erpclaw stripe-mrr-breakdown --account-id <id> --as-of 2026-04-30`
This shows the math line-by-line.

## Where to get help

- Email `support@avansaber.com` with your account-id and the error

- GitHub issues: github.com/avansaber/erpclaw-addons/issues

- Source code is public; the action implementations show exactly what runs

See Stripe features

---

## ERP comparison: 17 ERP and accounting tools side by side
URL: https://www.erpclaw.ai/erp-comparison/

# 17 ERP and accounting tools, side by side
License, self-hosting, AI architecture, scope, and pricing for 17 ERP and accounting tools. Honest, structured, sourced. Use it to short-list.
Updated 2026-05-09. Pricing reflects published rates; founder-reported quotes flagged inline.
Prefer it as a PDF? Grab the ERPClaw vs NetSuite vs Rillet comparison PDF, free.

## Quick read
Of 17 tools, only 2 are AI-native by architecture: ERPClaw and Rillet. ERPClaw is the only one that is also open-source AND self-hosted AND covers full ERP scope. Rillet is closed-source SaaS, finance-close-focused. Every other tool is AI-decorated: a chat box on a forms-and-workflows product.

- ·Open source AND AI-native: ERPClaw only.
- ·Self-hosted: ERPClaw, ERPNext, Odoo Community.
- ·$0 forever (no subscription): ERPClaw, ERPNext, Odoo Community.
- ·Full ERP scope: ERPClaw, NetSuite, Odoo, ERPNext.
- ·AI-native by design: ERPClaw, Rillet.

## Full matrix
| Tool | Category | License | Self-host | AI posture | Scope | Pricing
| ERPClaw | Open-source AI-native ERP | Free &amp; Open Source | Yes (SQLite or PostgreSQL via PyPika) | AI-native | Full ERP (every module, 14 industries) | $0 forever
| NetSuite | Enterprise SaaS ERP | Proprietary | No (Oracle cloud) | AI-decorated (NetSuite AI) | Full ERP | $10K+/year + implementation $50K-500K
| Sage Intacct | Mid-market financial SaaS | Proprietary | No | AI-decorated (Sage Copilot) | Financials + accounting | $15K+/year + 6-figure implementation
| Rillet | AI-native finance SaaS | Proprietary | No (vendor cloud) | AI-native | Finance close (not full ERP) | Seat-based, founder-reported $2K-10K/mo
| Odoo | Open-core ERP | LGPL (Community) + proprietary tiers | Yes (Community); No (Online/Enterprise) | AI-decorated (plug-in stack) | Full ERP (paid Studio gates much of value) | Community free; Standard $31.10/user/mo; Custom $61/user/mo
| ERPNext | Open-source ERP | Free &amp; Open Source | Yes | AI-decorated (changAI, NextAI, Ollama, Composio) | Full ERP | $0 (self-host) or Frappe Cloud SaaS
| QuickBooks | SMB accounting SaaS | Proprietary | No | AI-decorated (Intuit Assist) | Accounting only | QBO Plus $115/mo (capped at 5 users)
| Xero | SMB accounting SaaS | Proprietary | No | AI-decorated (Just Ask Xero) | Accounting only | Established plan $80/mo (US)
| FreshBooks | Freelancer / agency accounting | Proprietary | No | AI-decorated (FreshBooks AI) | Time-billing + light accounting | $30 to $60/mo + per-team-member fees
| A2X | Shopify-Amazon accounting connector | Proprietary | No | Rules-based (no AI claim) | Sync only | $19 to $179/mo per channel
| Synder | Multi-channel accounting connector | Proprietary | No | AI-decorated (Synder AI) | Sync + light reconciliation | $48 to $275/mo
| Bookkeep | Daily-summary accounting connector | Proprietary | No | Rules-based | Sync only (Shopify, Stripe, Square) | $70 to $295/mo
| Webgility | E-commerce accounting connector | Proprietary | No | Rules-based | Sync only (multi-channel) | $59 to $499/mo
| Acumatica | Mid-market cloud ERP | Proprietary | Yes (private cloud) | AI-decorated | Full ERP | Resource-based, $50K-$300K typical all-in
| Microsoft Dynamics 365 | Per-app enterprise ERP/CRM | Proprietary | Mostly cloud | AI-decorated (Copilot) | Full ERP via per-app licensing | Business Central $80-$110/user/mo; F&O $210+ (Oct 2025 raise)
| SAP Business One | SMB SAP ERP | Proprietary | Yes (on-prem) or cloud | AI-decorated (partner add-ons; Joule is for S/4HANA, not B1) | Full ERP | $95-$219/user/mo Professional + $1.4K-$3.5K/user one-time license + 18-20% maintenance
| Workday | Enterprise HCM + Financials | Proprietary | No (cloud only) | AI-decorated (Workday Illuminate) | HCM + Financials (Fortune 500 tier) | $100-$200/employee/year (founder-reported)

## Full matrix

### ERPClaw
Open-source AI-native ERP
LicenseFree &amp; Open Source
Self-hostYes (SQLite or PostgreSQL via PyPika)
AI postureAI-native
ScopeFull ERP (every module, 14 industries)
Pricing$0 forever

### NetSuite
Enterprise SaaS ERP
LicenseProprietary
Self-hostNo (Oracle cloud)
AI postureAI-decorated (NetSuite AI)
ScopeFull ERP
Pricing$10K+/year + implementation $50K-500K

### Sage Intacct
Mid-market financial SaaS
LicenseProprietary
Self-hostNo
AI postureAI-decorated (Sage Copilot)
ScopeFinancials + accounting
Pricing$15K+/year + 6-figure implementation

### Rillet
AI-native finance SaaS
LicenseProprietary
Self-hostNo (vendor cloud)
AI postureAI-native
ScopeFinance close (not full ERP)
PricingSeat-based, founder-reported $2K-10K/mo

### Odoo
Open-core ERP
LicenseLGPL (Community) + proprietary tiers
Self-hostYes (Community); No (Online/Enterprise)
AI postureAI-decorated (plug-in stack)
ScopeFull ERP (paid Studio gates much of value)
PricingCommunity free; Standard $31.10/user/mo; Custom $61/user/mo

### ERPNext
Open-source ERP
LicenseFree &amp; Open Source
Self-hostYes
AI postureAI-decorated (changAI, NextAI, Ollama, Composio)
ScopeFull ERP
Pricing$0 (self-host) or Frappe Cloud SaaS

### QuickBooks
SMB accounting SaaS
LicenseProprietary
Self-hostNo
AI postureAI-decorated (Intuit Assist)
ScopeAccounting only
PricingQBO Plus $115/mo (capped at 5 users)

### Xero
SMB accounting SaaS
LicenseProprietary
Self-hostNo
AI postureAI-decorated (Just Ask Xero)
ScopeAccounting only
PricingEstablished plan $80/mo (US)

### FreshBooks
Freelancer / agency accounting
LicenseProprietary
Self-hostNo
AI postureAI-decorated (FreshBooks AI)
ScopeTime-billing + light accounting
Pricing$30 to $60/mo + per-team-member fees

### A2X
Shopify-Amazon accounting connector
LicenseProprietary
Self-hostNo
AI postureRules-based (no AI claim)
ScopeSync only
Pricing$19 to $179/mo per channel

### Synder
Multi-channel accounting connector
LicenseProprietary
Self-hostNo
AI postureAI-decorated (Synder AI)
ScopeSync + light reconciliation
Pricing$48 to $275/mo

### Bookkeep
Daily-summary accounting connector
LicenseProprietary
Self-hostNo
AI postureRules-based
ScopeSync only (Shopify, Stripe, Square)
Pricing$70 to $295/mo

### Webgility
E-commerce accounting connector
LicenseProprietary
Self-hostNo
AI postureRules-based
ScopeSync only (multi-channel)
Pricing$59 to $499/mo

### Acumatica
Mid-market cloud ERP
LicenseProprietary
Self-hostYes (private cloud)
AI postureAI-decorated
ScopeFull ERP
PricingResource-based, $50K-$300K typical all-in

### Microsoft Dynamics 365
Per-app enterprise ERP/CRM
LicenseProprietary
Self-hostMostly cloud
AI postureAI-decorated (Copilot)
ScopeFull ERP via per-app licensing
PricingBusiness Central $80-$110/user/mo; F&O $210+ (Oct 2025 raise)

### SAP Business One
SMB SAP ERP
LicenseProprietary
Self-hostYes (on-prem) or cloud
AI postureAI-decorated (partner add-ons; Joule is for S/4HANA, not B1)
ScopeFull ERP
Pricing$95-$219/user/mo Professional + $1.4K-$3.5K/user one-time license + 18-20% maintenance

### Workday
Enterprise HCM + Financials
LicenseProprietary
Self-hostNo (cloud only)
AI postureAI-decorated (Workday Illuminate)
ScopeHCM + Financials (Fortune 500 tier)
Pricing$100-$200/employee/year (founder-reported)

Comparison directory

## Read every ERP comparison
Open a focused comparison, review the market matrix, or move into a migration guide for the system you use today.
QuickBooksNetSuiteSage IntacctRilletOdooOdoo Community and EnterpriseERPNextERPNext stack complexityXeroAcumaticaMicrosoft DynamicsSAP Business OneSAP Business One architectureWorkdayFreshBooksSynderA2XBookkeepWebgilityCampfireDualEntryPuzzleAI ERP stackBillowMigration guidesPlan a move from your current systemAI for QuickBooksEvaluate the QuickBooks transition pathAI for NetSuiteKeep and augment NetSuite or plan a move

## Methodology
License + self-hosting + scope are read from the vendor's public docs as of 2026-05-09. AI posture follows the AI-native vs AI-decorated framework (see AI-native ERP): a tool is AI-native if AI is the primary design surface from line one, AI-decorated if AI is a chat sidebar added to a forms-and-workflows product after launch.
Pricing reflects publicly-listed plans where available. Where a vendor's pricing is "contact sales", we cite founder-reported quotes (Rillet) or industry-standard ranges (NetSuite). All prices in USD.
Sources: vendor pricing pages, vendor product pages, G2 + Capterra reviews, public ChatGPT GPT-5 conversations on the AI-native category (captured 2026-05-05). Per-vendor deep-dive pages link from each row.

Related: read the AI-native ERP framework that classifies the AI-posture column, the open-source AI accounting map (the open-source slice), the AI accounting story, the ERPClaw pricing page, or the feature overview.
Long-form deep dives: NetSuite alternative, Sage Intacct alternative, ERPNext vs ERPClaw, Odoo alternative, Xero alternative, FreshBooks alternative, and the 5 AI-native ERPs that earn the label.

---

## Features - Every module of AI-Native ERP, Free
URL: https://www.erpclaw.ai/features/

# Every module your
business needs
ERPClaw covers accounting, inventory, manufacturing, HR, CRM, regional compliance, and 14 industry verticals, all fully integrated. All operated through natural language. Plus ERPClaw OS, the self-improving engine.

Tier 1

## Foundation

Explore this tier

### Setup

Company creation, chart of accounts, fiscal years, cost centers, currency, seed data, RBAC, database backup, restore, and encryption.

### General Ledger

Double-entry GL posting, period closing, trial balance queries, account management, naming series, GL integrity check, and immutable audit trail.

Explore all actions

Tier 2

## Financial Operations

Explore this tier

### Journals

Journal entries with multi-line debit/credit, reversals, recurring journals, intercompany transactions, and recurring templates.

### Payments

Payment entries, bank reconciliation, automatic allocation against invoices, payment terms, advance payments, and early payment discounts.

### Tax

US sales tax templates, tax rules, withholding tax, 1099 tracking, tax categories, and automatic tax calculation.

### Reports

Trial balance, P&L, balance sheet, cash flow, general ledger report, accounts receivable/payable aging, budget vs actuals, and elimination rules.

Explore all actions

Tier 3

## Supply Chain

Explore this tier

### Inventory

Item master, warehouses, stock entries, batch tracking, serial numbers, pricing rules, stock reconciliation, revaluation, and CSV import. Warehouse depth: hard stock reservations, putaway routing, guided pick lists that become delivery notes, and ranked item substitutes that manufacturing inherits.

### Selling

Full order-to-cash: customers, quotations, sales orders, delivery notes, sales invoices, credit notes, recurring invoices, intercompany mirroring.

### Buying

Procure-to-pay: suppliers, material requests, RFQs, purchase orders, purchase receipts, purchase invoices, debit notes, landed costs.

### Manufacturing

Bill of materials, work orders, job cards, production planning (MRP), operation tracking, scrap management, subcontracting.

Explore all actions

Tier 4

## Enterprise

Explore this tier

### HR

Employee master, departments, designations, leave management, attendance, expense claims, and employee lifecycle events.

### Payroll

US payroll: salary structures, components, payroll runs, salary slips, FICA, federal/state income tax withholding, W-2 generation.

### Projects

Projects, tasks, timesheets, Gantt dependencies, budgets, project profitability, milestones, and resource allocation.

### Assets

Fixed asset register, depreciation schedules (straight-line and declining balance), asset movement, disposal, and repair tracking.

### Quality

Quality inspections, non-conformance reports, corrective/preventive actions, inspection templates, and quality dashboards.

Explore all actions

Tier 5

## Extended Modules

Explore this tier

### CRM

Built-in Sales: run the whole pipeline from first lead to closed order, with contacts and companies, follow-up tasks, custom deal stages, saved views, search across every sales record, and CSV import and export, all in the same system as your books.

### Support

Support issues, SLA management, warranty claims, maintenance scheduling, service contracts, and escalation rules.

### Billing

Usage-based metering, rate plans, bill runs, prepaid credits, billing adjustments, and automatic invoice generation.

### AI Engine

Anomaly detection, cash flow forecasting, predictive scoring, business rules, relationship scoring, and conversation memory.

### Analytics

Cross-module KPIs, financial ratios, dashboards, revenue/expense trends, inventory analytics, HR metrics, and period comparisons.

Explore all actions

Tier 6

## Integrations & Regional

Explore this tier

### Plaid

Bank account integration via Plaid: link accounts, sync transactions, and auto-match with GL entries.

### Stripe

Full Stripe integration end to end: account setup, sync engine, customer mapping, GL rules, reconciliation, GL posting, and reporting. Auto-reconciliation with 3-layer matching and GL posting for charges, refunds, disputes, and payouts.

### S3 Backup

Cloud backup to Amazon S3: upload, list, restore, and manage remote database backups.

### Region: India

GST (post GST 2.0), e-invoicing, GSTR-1/3B, TDS, Indian CoA (Ind-AS), PF/ESI/PT payroll deductions, and ID validation.

### Region: Canada

GST/HST/PST/QST, CPP/CPP2/QPP/EI, federal + provincial income tax, T4/T4A/ROE/PD7A, Canadian CoA (ASPE), BN/SIN validation.

### Region: UK

VAT (standard/reduced/zero/flat rate), PAYE, NI, student loan, pension (NEST), RTI (FPS/EPS/P60/P45), CIS, FRS 102 CoA.

### Region: EU

VAT (27 member states), reverse charge, OSS, Intrastat, EN 16931 e-invoicing, SAF-T, EC Sales List, VIES, withholding tax.

Explore all actions

Tier 7

## ERPClaw OS

Explore this tier

### Constitutional Framework

A constitutional framework governs every operation: financial integrity, data ownership, immutability rules, and safety constraints that can never be overridden.

### Module Generation

Deterministic generation patterns. Create new modules from proven templates with automatic schema migration, test generation, and sandbox validation.

### Semantic Correctness Engine

Validates that every change preserves system invariants. Cross-references actions against constitutional articles before deployment.

### Self-Improvement Log

Tracks every improvement, heartbeat analysis per module, and DGM variant engine for evolutionary optimization of module patterns.

Explore all actions

Industry Verticals

## Purpose-Built Solutions
14 industry verticals extend ERPClaw with domain-specific workflows, compliance, and reporting.

### PropClaw
Real estate
Property management: properties, leases, tenants, maintenance, and accounting. Purpose-built for the vertical, not bolted on.

### HealthClaw
Healthcare
Healthcare with 5 sub-verticals for dental, mental health, veterinary, and home health. HIPAA-compliant workflows.

### EduClaw
Education
Education management covering: core, financial aid, K-12, scheduling, LMS, state reporting, and higher ed. FERPA compliant.

Included Free

## Web Dashboard (Beta)
Every skill works in the chat AND in the browser. Webclaw auto-generates forms, data tables, and dashboards from skill definitions, zero manual config.

#### Dynamic Forms
Entity lookups, date pickers, and child tables, all auto-generated from SKILL.md definitions.

#### Data Tables
Search, sort, filter, and paginate any entity across all skills. CSV export included.

#### AI Chat Panel
Ask questions about your data while browsing any skill. Context-aware responses.

Dashboard

Chart of Accounts

Procurement

Mobile

Capability directory

---

## Dimensional Reporting: Profit by Department or Project
URL: https://www.erpclaw.ai/features/dimensional-reporting/

# Which part of the business actually makes the money?
One total hides everything. ERPClaw lets every entry carry the labels you run on, department, project, location, anything, and then answers along those lines: profit by project, costs by team, margin by store. Asked in plain language, answered from the books.
How it worksGet ERPClaw free

## Ask along the lines you manage
"Show me this month's profit broken down by department."
Revenue, costs, and net for each department, side by side, from the books.

"What did the Henderson project actually make us?"
Everything tagged to that project, income and costs together, one honest number.

"Which store is carrying the other one?"
The same report, split by location instead, without rebuilding anything.

## Common questions

### What is dimensional reporting, in plain terms?
It means every entry in your books can carry labels like department, project, location, or campaign, and every report can split by them. 'Profit by project' stops being a spreadsheet afternoon and becomes a question.

### How is this different from categories in my accounting app?
Categories tell you what money was, like rent or sales. Dimensions tell you whose it was: which team, which job, which site. You need both to know which parts of the business actually earn.

### Do I have to restructure my books?
No. You label entries as you make them, in plain language: 'book 120 dollars of supplies against Engineering.' The label travels with the entry from then on, and untagged history simply shows in its own bucket.

### Can I invent my own labels?
Yes. Department, project, and cost center are ready out of the box, and you can add the ones your business actually runs on: region, fund, campaign, store. Retiring one is protected so you cannot orphan live data by accident.

### Which reports can split this way?
Profit and loss grouped by any dimension, and the trial balance, balance sheet, cash flow, and ledger narrowed to one. Scope to a project and split by department in the same request.

### Is this a paid tier?
No. It is part of the free, self-hosted, open-source ERPClaw, like everything else.

Related: audit-ready books · ERPClaw Insights, your AI analyst · all features

---

## Enterprise ERP - HR, Payroll, Projects, Assets, Quality
URL: https://www.erpclaw.ai/features/enterprise/

# Enterprise operations
Five skills for people management, payroll, project tracking, asset lifecycle, and quality control. Run HR, cut paychecks, and track depreciation, all from chat.
HR (28)Payroll (18)Projects (18)Assets (16)Quality (14)
Webclaw 3-panel layout, sidebar, data table, and AI chat

## HR
Full coverage
Employee master data, departments, designations, leave management, attendance tracking, expense claims, and employee lifecycle events.

### Employees
`add-employee`Create a new employee record
`update-employee`Modify employee details
`get-employee`Retrieve employee information
`list-employees`Browse all employees with filters
`inactivate-employee`Mark employee as inactive

### Organization
`add-department`Create a new department
`add-designation`Create a job title / designation
`add-employee-grade`Define an employee grade level
`list-departments`Browse all departments

### Leave
`add-leave-type`Define a leave type (PTO, sick, etc.)
`add-leave-allocation`Allocate leave days to employee
`add-leave-application`Submit a leave request
`approve-leave-application`Approve or reject leave request
`get-leave-balance`Check remaining leave balance
`list-leave-applications`Browse all leave requests

### Attendance
`add-attendance`Record daily attendance
`list-attendance`Browse attendance records

### Expenses
`add-expense-claim`Submit an expense claim
`approve-expense-claim`Approve or reject expense claim
`submit-expense-claim`Post expense claim to GL

### Lifecycle
`record-lifecycle-event`Record promotion, transfer, etc.
`list-lifecycle-events`Browse employee lifecycle history
`add-holiday-list`Create a holiday calendar

### System
`list-holidays`Browse all holidays
`status`Check skill health status

### Example prompts
"Add employee John Smith, Engineering, $85,000""John is requesting 3 days PTO""Submit expense claim for hotel"

## Payroll
Full coverage
US payroll processing: salary structures, components, payroll runs, salary slips, FICA, federal and state income tax withholding, and W-2 generation.

### Structure
`add-salary-structure`Create a salary structure template
`update-salary-structure`Modify a salary structure
`get-salary-structure`Retrieve structure details
`list-salary-structures`Browse all salary structures

### Components
`add-salary-component`Define an earning or deduction
`list-salary-components`Browse all salary components
`add-income-tax-slab`Configure income tax brackets

### Assignments
`add-salary-assignment`Assign structure to employee
`list-salary-assignments`Browse all assignments

### Slips
`add-salary-slip`Create a salary slip draft
`update-salary-slip`Modify a draft salary slip
`get-salary-slip`Retrieve salary slip details
`list-salary-slips`Browse all salary slips
`submit-salary-slip`Post salary slip to GL

### Runs
`run-payroll`Process payroll for a period
`list-payroll-runs`Browse all payroll runs
`get-payroll-run`Retrieve payroll run details

### Tax
`generate-w2-data`Generate W-2 tax data for year-end

### System
`status`Check skill health status

### Example prompts
"Run payroll for February 2026""Show John's salary slip""Generate W-2 data for 2025"

## Projects
Full coverage
Project management with tasks, milestones, timesheets, budget tracking, resource utilization, and project profitability analysis.

### Projects
`add-project`Create a new project
`get-project`Retrieve project details
`list-projects`Browse all projects

### Tasks
`add-task`Create a task under a project
`update-task`Modify task details or status
`list-tasks`Browse all tasks with filters
`add-milestone`Define a project milestone

### Timesheets
`add-timesheet`Create a timesheet entry
`update-timesheet`Modify a draft timesheet
`get-timesheet`Retrieve timesheet details
`list-timesheets`Browse all timesheets
`submit-timesheet`Finalize and post timesheet

### Reports
`get-project-budget-tracking`Compare budget vs actual spend
`get-project-schedule`View project timeline and Gantt
`resource-utilization`Analyze resource allocation

### Other
`list-projects-by-customer`Filter projects by customer
`close-project`Mark project as complete
`status`Check skill health status

### Example prompts
"Create project Website Redesign""Log 8 hours on task""Show project budget vs actual"

## Assets
Full coverage
Fixed asset register, depreciation schedules (straight-line and declining balance), asset movements, maintenance tracking, and disposal.

### Assets
`add-asset`Register a new fixed asset
`get-asset`Retrieve asset details
`list-assets`Browse all assets with filters

### Categories
`add-asset-category`Define an asset category
`list-asset-categories`Browse asset categories

### Depreciation
`get-depreciation-schedule`View depreciation schedule
`post-depreciation`Post depreciation entries to GL
`depreciation-summary`Summary of depreciation by category
`update-useful-life`Revise asset useful life

### Movements
`add-asset-movement`Record asset location change
`add-asset-maintenance`Schedule or log maintenance
`add-asset-disposal`Create asset disposal record

### System
`submit-asset-disposal`Post disposal entries to GL
`list-asset-movements`Browse asset movement history
`list-asset-maintenance`Browse maintenance records
`status`Check skill health status

### Example prompts
"Add asset Laptop, $2,000, 3yr life""Run depreciation for January""Dispose of old server"

## Quality
Full coverage
Quality inspections with configurable templates, non-conformance tracking, quality goals, defect rate analysis, and quality dashboards.

### Inspection
`add-quality-inspection-template`Create an inspection template
`add-quality-inspection-parameter`Add parameter to template
`add-quality-inspection`Start a quality inspection
`add-quality-inspection-reading`Record an inspection reading
`list-quality-inspections`Browse all inspections

### Defects
`add-non-conformance`Report a quality defect
`add-quality-goal`Set a quality target
`get-non-conformance`Retrieve defect report details
`list-non-conformances`Browse all non-conformances

### Reports
`submit-quality-inspection`Finalize inspection with verdict
`quality-dashboard`View quality metrics overview
`get-defect-rate`Calculate defect rate percentage
`get-inspection-pass-rate`Calculate inspection pass rate
`status`Check skill health status

### Example prompts
"Create inspection for Batch-001""Record quality reading""Show defect rate"

## Enterprise-grade, chat-simple
Everything you need for HR, payroll, projects, assets, and quality. All included free.
Get Started Free

Related: read the architecture in AI-native ERP, see the self-improving engine, or compare against the entrenched mid-market choice at NetSuite. Pricing posture: $0 forever. Test surface: quality.

---

## ERPClaw OS: Generate ERP Modules From a Sentence
URL: https://www.erpclaw.ai/features/erpclaw-os/

# The self-improving ERP engine
Every ERP processes transactions. Only ERPClaw learns from them. constitutional articles. 3 evolution phases. 701 OS-specific tests. The only ERP that detects gaps and generates modules while protecting financial integrity.

## Constitutional Articles
Inviolable rules that govern every operation. The AI cannot override them. No admin can bypass them. They are the foundation of trust.
1Financial data integrity is inviolable

2Double-entry accounting is always enforced

3GL entries are immutable, cancel creates reversal

4All financial amounts use Decimal (TEXT storage)

5IDs use UUID4 (TEXT storage)

6Submit operations are atomic (single transaction)

712-step GL validation on every posting

8Only owning module can write its tables

9Any module can read any table

10Parameterized queries only, no string interpolation

11Output is always JSON to stdout

12Errors are JSON with exit code 1

13WAL mode and FK enforcement always ON

14Draft-to-Submit lifecycle on all documents

15Naming series follow deterministic patterns

16RBAC governs all data access

17AES-256 encryption for sensitive fields

18Backup is always recoverable

## Three Evolution Phases
ERPClaw OS grew through three phases, each adding more autonomy while maintaining safety.

### Phase 1: Child
255 tests

- Module generation from 7 deterministic patterns
- SKILL.md validation and linting
- Sandbox testing before deployment
- Schema DDL generation and migration
- Constitutional article compliance checks

### Phase 2: Teenager
226 tests

- Tier classification system (Tier 0-3 autonomy)
- Schema migration engine
- Deploy pipeline with safety gates
- Install-suite for grouped installs
- Adversarial audit and compliance weather

### Phase 3: Adult
220 tests

- Semantic correctness engine
- Self-improvement log with change tracking
- DGM variant engine (evolutionary optimization)
- Heartbeat analysis per module
- Gap detection across business workflows

## Safety Model
Self-improvement sounds risky. Here is how ERPClaw OS ensures every change is safe.

### 11 Protected Files
Core financial files (gl_posting.py, stock_posting.py, etc.) can NEVER be modified by the AI. Hard-coded exclusion list.

### Constitutional Tests
Automated tests verify every article is enforced. Run before every deployment.

### Tier Classification
Tier 0 (manual only) through Tier 3 (autonomous). Financial modules are always Tier 0.

### Invariant Checks
GL double-entry balance, voucher balance, immutability verification per transaction batch.

### Sandbox Validation
Every generated module runs in an isolated sandbox before deployment to production.

### Semantic Verification
Changes are validated against business rules, not just syntax. Catches logic errors before they reach production.

## See the code behind the OS
Every constitutional article, every safety check, every generation pattern, open source under open source license.
View on GitHub

Related: read the architecture in AI-native ERP, see the test surface at quality, the foundation features, or the engineering deep-dive in building ERPClaw with Claude Code.

---

## Extended ERP - CRM, Support, Billing, AI Engine
URL: https://www.erpclaw.ai/features/extended/

# Extended Modules
CRM, customer support, usage-based billing, AI-powered anomaly detection, and real-time analytics dashboards, the intelligence layer that turns data into decisions.

## CRM
CRM
Full-cycle customer relationship management. Capture leads from any source, track opportunities through your sales pipeline, run campaigns, and log every call, email, and meeting, all in natural language.

### All Actions
`add-lead`Capture a new lead from any source

`update-lead`Edit lead details or status

`get-lead`Retrieve full lead profile

`list-leads`Query leads with filters

`add-opportunity`Create a sales opportunity

`update-opportunity`Update opportunity stage or value

`get-opportunity`Retrieve opportunity details

`list-opportunities`List all open opportunities

`close-opportunity`Mark opportunity won or lost

`add-campaign`Create a marketing campaign

`add-campaign-lead`Assign a lead to a campaign

`list-campaigns`List all campaigns

`add-crm-activity`Schedule a CRM activity

`add-communication`Log a call, email, or meeting

`list-activities`List activities for a lead or deal

`pipeline-report`View sales pipeline summary

`list-crm-activities`List all CRM activity types

`status`Health check for the CRM skill

### Example Commands
`Add lead from trade show`
`Convert lead to opportunity`
`Show sales pipeline`
`Log a call with prospect`

## Support
Support
Track customer issues from first report to resolution. Define SLAs with priority-based response times, manage warranty claims, schedule preventive maintenance, and surface overdue tickets before they escalate.

### All Actions
`add-issue`Open a new support ticket

`update-issue`Update issue status or assignment

`get-issue`Retrieve full issue details

`list-issues`Query issues with filters

`close-issue`Resolve and close a ticket

`add-issue-comment`Add a comment to an issue

`list-issue-comments`List all comments on an issue

`add-sla`Define a service level agreement

`add-sla-priority`Set response/resolution targets

`get-sla-status`Check SLA compliance for an issue

`list-slas`List all SLA definitions

`add-warranty-claim`File a warranty claim

`get-warranty-claim`Retrieve warranty claim details

`list-warranty-claims`List all warranty claims

`add-maintenance-schedule`Create a maintenance schedule

`add-maintenance-visit`Log a completed maintenance visit

`list-maintenance-schedules`List all maintenance schedules

`overdue-issues-report`Show overdue and at-risk tickets

`status`Health check for the Support skill

### Example Commands
`Create support ticket for Acme`
`Check SLA status`
`Show overdue issues`
`Add warranty claim`

## Billing
Billing
Usage-based and subscription billing made simple. Define meters and rate plans, record usage events, calculate billable amounts with tiered pricing, manage prepaid balances, and generate invoices automatically.

### All Actions
`add-meter`Define a usage meter (e.g. API calls)

`add-meter-reading`Record a meter reading

`list-meter-readings`List readings for a meter

`add-rate-plan`Create a pricing rate plan

`add-rate-tier`Add a tier to a rate plan

`list-rate-plans`List all rate plans

`add-usage-event`Log a billable usage event

`create-billing-period`Open a new billing period

`calculate-billable-usage`Calculate charges for a period

`create-billing-adjustment`Apply a billing adjustment

`get-customer-prepaid-balance`Check customer prepaid balance

`create-invoice-from-usage`Generate invoice from usage data

`list-billing-periods`List all billing periods

`submit-billing-invoice`Submit and finalize a billing invoice

`record-payment-against-prepaid`Record payment to prepaid account

`add-prepaid-credit`Add prepaid credit to a customer

`get-prepaid-balance`Retrieve current prepaid balance

`refund-prepaid-credit`Refund unused prepaid credit

`status`Health check for the Billing skill

### Example Commands
`Add meter for API usage`
`Record meter reading 1,500 calls`
`Run billing for January`
`Check prepaid balance`

## AI Engine
AI Engine
Built-in intelligence for your ERP. Detect anomalies in financial data, forecast cash flow, evaluate business rules automatically, score vendor and customer relationships, and maintain full conversational audit trails.

### All Actions
`add-anomaly`Flag a detected anomaly

`list-anomalies`List all flagged anomalies

`confirm-anomaly`Confirm or dismiss an anomaly

`add-cash-flow-forecast`Generate a cash flow forecast

`get-cash-flow-forecast`Retrieve a stored forecast

`add-correlation-analysis`Run a correlation analysis

`list-correlations`List all correlation analyses

`add-business-rule`Define an automated business rule

`evaluate-business-rules`Run all rules against current data

`add-pending-decision`Queue a decision for review

`list-pending-decisions`List all pending decisions

`add-categorization-rule`Create a categorization rule

`add-relationship-score`Score a vendor or customer

`get-relationship-score`Retrieve a relationship score

`update-relationship-score`Recalculate a relationship score

`save-conversation-context`Persist conversation context

`get-conversation-context`Retrieve saved context

`log-audit-conversation`Log conversation for audit trail

`status`Health check for the AI Engine skill

### Example Commands
`Detect anomalies in cash flow`
`Forecast next 6 months`
`Score vendor reliability`
`Evaluate business rules`

## Analytics
Analytics
Real-time KPIs, dashboards, and trend analysis across your entire business. Track financial ratios, operational metrics, and HR indicators. Compare periods, analyze budget variance, and define custom KPIs, all queryable in plain English.

### All Actions
`get-revenue-kpi`Current revenue metrics

`get-expense-kpi`Current expense metrics

`get-profit-margin`Gross and net profit margins

`get-current-ratio`Current assets / current liabilities

`get-quick-ratio`Quick (acid-test) ratio

`get-debt-to-equity`Debt-to-equity ratio

`get-inventory-turnover`Inventory turnover rate

`get-days-sales-outstanding`Average collection period

`get-days-payable-outstanding`Average payment period

`get-working-capital`Net working capital

`get-employee-count`Total active employees

`get-attrition-rate`Employee attrition rate

`get-avg-salary`Average salary across org

`get-revenue-trend`Revenue trend over time

`get-expense-trend`Expense trend over time

`get-cash-flow-trend`Cash flow trend over time

`get-financial-dashboard`Full financial KPI dashboard

`get-operational-dashboard`Operational metrics dashboard

`get-hr-dashboard`HR metrics dashboard

`get-period-comparison`Compare two periods side by side

`get-budget-variance-analysis`Budget vs actual variance

`add-custom-kpi`Define a custom KPI formula

`get-custom-kpi`Calculate a custom KPI

`status`Health check for the Analytics skill

### Example Commands
`Show financial dashboard`
`What's our profit margin?`
`Compare this quarter vs last`
`Days sales outstanding`

## Intelligence built in, not bolted on
CRM, support, billing, anomaly detection, and analytics, all working together through natural language. Self-host free.
Get Started Free Back to All Features

Related: see the architecture in AI-native ERP, the foundation tier that the extended modules build on, the AI accounting story, or comparisons against NetSuite and Odoo.

---

## Accounting, GL, Payments and Tax in One System
URL: https://www.erpclaw.ai/features/financial-ops/

# Financial Operations
Day-to-day financial workflows. Journal entries, payment processing, tax calculation with 1099 compliance, and a full suite of financial reports, all through conversational commands.
General Ledger

Financial Dashboard

## Journals
Journals
Full journal entry lifecycle with multi-line debit/credit support. Handle opening entries, closing entries, depreciation, exchange rate revaluations, write-offs, intercompany transactions, and credit/debit notes, all with draft-to-submit workflow.

### Entry Types
journalopeningclosingdepreciationexchange_rate_revaluationwrite_offinter_companycredit_notedebit_note

### All Actions
`add-journal-entry`Create a new journal entry draft

`update-journal-entry`Modify a draft journal entry

`get-journal-entry`Retrieve a journal entry by ID

`list-journal-entries`List journal entries with filters

`submit-journal-entry`Validate and post to the GL

`cancel-journal-entry`Reverse a submitted entry

`amend-journal-entry`Amend a cancelled entry

`delete-journal-entry`Delete a draft entry permanently

`duplicate-journal-entry`Copy an entry as a new draft

`status`Health check for the Journals skill

### Example Commands
`Create a journal entry: debit Office Supplies $500, credit Cash $500`
`Submit journal entry JE-2026-00001`

## Payments
Payments
End-to-end payment processing and bank reconciliation. Record customer receipts and supplier payments, allocate against outstanding invoices, reconcile bank statements, and track unallocated balances, everything needed for cash management.

### All Actions
`add-payment`Record a new payment entry draft

`update-payment`Modify a draft payment

`get-payment`Retrieve payment details by ID

`list-payments`List payments with date/party filters

`submit-payment`Validate and post payment to GL

`cancel-payment`Reverse a submitted payment

`delete-payment`Delete a draft payment permanently

`allocate-payment`Allocate payment against invoices

`reconcile-payments`Auto-match payments to bank txns

`bank-reconciliation`Run full bank statement reconciliation

`get-unallocated-payments`List payments with open balances

`create-payment-ledger-entry`Create a payment ledger record

`get-outstanding`Show outstanding amount for a party

`status`Health check for the Payments skill

### Example Commands
`Record payment of $1,200 from Acme Corp`
`Reconcile bank statement`
`Show unallocated payments`

## Tax
Tax
US-focused tax engine with templates, rules, and automatic calculation. Configure sales tax rates, set up tax categories and rules for automatic resolution, handle withholding tax, track 1099 payments, and generate year-end 1099 data, fully compliant.

### All Actions
`add-tax-template`Create a tax rate template

`update-tax-template`Modify a tax template

`get-tax-template`Retrieve tax template details

`list-tax-templates`List all tax templates

`delete-tax-template`Remove a tax template

`resolve-tax-template`Auto-resolve template for a transaction

`calculate-tax`Calculate tax for a line item

`add-tax-category`Create a tax category grouping

`list-tax-categories`List all tax categories

`add-tax-rule`Define a conditional tax rule

`list-tax-rules`List all active tax rules

`add-item-tax-template`Assign tax template to an item

`add-tax-withholding-category`Create a withholding tax category

`get-withholding-details`Look up withholding rates for a party

`record-withholding-entry`Record a tax withholding transaction

`record-1099-payment`Flag a payment as 1099-reportable

`generate-1099-data`Generate 1099 data for a tax year

`status`Health check for the Tax skill

### Example Commands
`Set up sales tax at 8.25%`
`Calculate tax for invoice`
`Generate 1099 data for 2025`

## Reports
Reports
Complete financial reporting suite. Generate trial balances, profit and loss statements, balance sheets, cash flow reports, aging schedules, budget variance analysis, and comparative period reports, all from a single natural-language request.

### All Actions
`trial-balance`Generate trial balance for a period

`profit-and-loss`Profit & loss statement by date range

`balance-sheet`Balance sheet as of a given date

`cash-flow`Cash flow statement for a period

`general-ledger`Detailed GL report with filters

`party-ledger`Ledger for a specific customer/supplier

`tax-summary`Tax collected and paid summary

`ar-aging`Accounts receivable aging schedule

`ap-aging`Accounts payable aging schedule

`budget-vs-actual`Budget vs actual variance report

`comparative-pl`Side-by-side P&L across periods

`payment-summary`Payment summary by party or method

`gl-summary`Summarized GL by account group

`status`Health check for the Reports skill

### Example Commands
`Show the trial balance`
`Generate P&L for January`
`Run accounts receivable aging`

---

## ERPClaw Foundation: Accounting, GL, Multi-Entity Core
URL: https://www.erpclaw.ai/features/foundation/

# Foundation
The bedrock of your ERP. Company setup, user management, chart of accounts, double-entry general ledger, fiscal year control, and immutable audit trails, everything starts here.
Chart of Accounts in the General Ledger module

## Setup
Setup
Bootstrap your entire business in minutes. Create companies, configure currencies, define payment terms, manage users with role-based access, and seed default data, all through natural language.

### All Actions
`setup-company`Initialize a new company with defaults

`update-company`Modify company settings and details

`get-company`Retrieve full company profile

`list-companies`Show all configured companies

`add-currency`Register a new currency (e.g. EUR, GBP)

`list-currencies`List all active currencies

`add-exchange-rate`Set a currency exchange rate

`get-exchange-rate`Look up current exchange rate

`list-exchange-rates`Show all exchange rate history

`add-payment-terms`Define payment terms template

`list-payment-terms`List all payment terms

`add-uom`Add a unit of measure (e.g. kg, box)

`list-uoms`List all units of measure

`add-uom-conversion`Define UoM conversion factor

`create-user`Create a new user account

`update-user`Update user profile or permissions

`list-users`List all users in the system

`get-user`Retrieve a single user profile

`seed-defaults`Populate default data and templates

`backup-database`Create a full database backup

### Example Commands
`Set up a company called Acme Corp`
`Add currency EUR`
`Create user john@example.com`
`Backup the database`

## General Ledger
General Ledger
Rock-solid double-entry accounting with immutable audit trails. Manage your chart of accounts, post and reverse GL entries, control fiscal periods, track budgets by cost center, and verify ledger integrity at any time.

### All Actions
`setup-chart-of-accounts`Import a standard chart of accounts

`add-account`Create a new GL account

`update-account`Modify account name or properties

`list-accounts`List all accounts in the COA

`get-account`Retrieve a single account detail

`freeze-account`Lock an account from further posting

`unfreeze-account`Unlock a frozen account

`post-gl-entries`Post validated double-entry GL lines

`reverse-gl-entries`Reverse a posted GL entry set

`list-gl-entries`Query GL entries with filters

`check-gl-integrity`Verify debits equal credits globally

`add-fiscal-year`Define a new fiscal year period

`list-fiscal-years`List all fiscal years

`validate-period-close`Pre-check before closing a period

`close-fiscal-year`Close a fiscal year permanently

`reopen-fiscal-year`Reopen a previously closed year

`add-cost-center`Create a cost center for tracking

`list-cost-centers`List all cost centers

`add-budget`Set a budget for account + cost center

`list-budgets`List all budgets and utilization

`seed-naming-series`Initialize document naming counters

`next-series`Generate next number in a series

`get-account-balance`Get current balance of any account

`status`Health check for the GL skill

### Example Commands
`Set up US GAAP chart of accounts`
`Show trial balance`
`Close fiscal year 2025`
`Check GL integrity`

---

## Integrations: Plaid, Stripe, Shopify, S3
URL: https://www.erpclaw.ai/features/integrations/

# Integrations & Regional
Connect to Plaid, Stripe, and S3 out of the box. Expand globally with region-specific tax, payroll, and compliance modules for India, Canada, the UK, and the EU.

## Plaid
Plaid
Connect your bank accounts securely through Plaid. Sync transactions automatically, match them to existing payments and invoices, and keep your books in sync with your bank, zero manual data entry.

### All Actions
`configure-plaid`Set up Plaid API credentials

`link-bank-account`Link a bank account via Plaid

`sync-bank-transactions`Pull latest transactions from bank

`match-transactions`Auto-match bank txns to ERP entries

`list-bank-transactions`List synced bank transactions

`unlink-bank-account`Disconnect a linked bank account

`status`Health check for the Plaid skill

### Example Commands
`Link my Chase account`
`Sync bank transactions`
`Auto-match payments`

## Stripe
Stripe
Full Stripe integration covering every operation across 7 domains. Connects your Stripe account to ERPClaw with auto-reconciliation (3-layer matching), GL posting for charges, refunds, disputes, and payouts, customer mapping, a sync engine, and full reporting. Everything lives in one schema and is covered by the regression suite.

### All Actions
`stripe-setup-account`Connect and configure Stripe account

`stripe-sync-transactions`Pull and sync Stripe transactions

`stripe-map-customer`Map Stripe customers to ERP contacts

`stripe-set-gl-rules`Configure GL posting rules for Stripe

`stripe-reconcile`Auto-reconcile with 3-layer matching

`stripe-post-gl`Post Stripe transactions to GL

`stripe-browse-transactions`Browse and filter Stripe data

`stripe-connect-account`Manage Stripe Connect sub-accounts

`stripe-report`Generate Stripe reconciliation reports

### Example Commands
`Connect my Stripe account`
`Reconcile Stripe payments`
`Post Stripe charges to GL`
`Show Stripe payout report`

## Shopify
Shopify
Full Shopify integration covering every operation: order sync, payout reconciliation, COGS tracking, gift card accounting, dispute handling, and GL posting. The clearing account pattern tracks every dollar in transit. Everything lives in one schema and is covered by the regression suite.

### All Actions
`shopify-setup-store`Connect and configure Shopify store

`shopify-sync-orders`Pull and sync Shopify orders

`shopify-sync-payouts`Sync Shopify payouts to clearing account

`shopify-set-gl-rules`Configure GL posting rules for Shopify

`shopify-reconcile`Reconcile clearing account against payouts

`shopify-post-gl`Post Shopify transactions to GL

`shopify-track-cogs`Track COGS per order and location

`shopify-gift-cards`Manage gift card deferred revenue

`shopify-report`Generate Shopify reconciliation reports

### Example Commands
`Connect my Shopify store`
`Reconcile Shopify payouts`
`Show COGS by product category`
`Show Shopify clearing account balance`

## S3 Backup
S3 Backup
Encrypted cloud backups to Amazon S3. Schedule automatic backups, browse your backup history, and restore from any point in time, your data is always safe, even if your server isn't.

### All Actions
`configure-s3`Set up S3 bucket and credentials

`upload-backup`Upload a database backup to S3

`list-remote-backups`List all backups stored in S3

`restore-from-s3`Restore database from an S3 backup

`delete-remote-backup`Delete a remote backup from S3

`get-backup-status`Check status of last backup job

`status`Health check for the S3 Backup skill

### Example Commands
`Backup database to S3`
`List remote backups`
`Restore from latest backup`

4 Regions

## Regional Compliance
Each regional module adds country-specific tax rules, payroll deductions, statutory reports, chart of accounts, and ID validation. Drop in only the regions you need.

### India
India

- GST: CGST, SGST, IGST calculation
- GSTR-1 and GSTR-3B generation
- E-invoicing and e-waybill
- TDS categories and returns
- Indian CoA (Ind-AS compliant)
- PF, ESI, Professional Tax
- Form 16 generation
- PAN, GSTIN, Aadhaar validation

### Canada
Canada

- GST / HST / PST / QST calculation
- CPP, CPP2, QPP, EI deductions
- Federal + provincial income tax
- T4, T4A, ROE, PD7A generation
- Canadian CoA (ASPE compliant)
- BN and SIN validation
- Multi-province payroll
- CRA-ready reporting

### United Kingdom
United Kingdom

- VAT: standard, reduced, zero, flat rate
- PAYE and National Insurance
- Student loan deductions
- Pension auto-enrolment (NEST)
- RTI: FPS, EPS, P60, P45
- CIS (Construction Industry Scheme)
- FRS 102 chart of accounts
- Making Tax Digital ready

### European Union
European Union

- VAT across 27 member states
- Reverse charge mechanism
- One Stop Shop (OSS)
- Intrastat reporting
- EN 16931 e-invoicing
- SAF-T audit file generation
- EC Sales List and VIES validation
- Withholding tax support

## Deep integration pages
Moving from another system, or wiring up your bank? These guides go deeper.
QuickBooks importNetSuite importBank feedsXero

## Connect everything. Comply everywhere.
Bank feeds, payment processing, cloud backups, and regional compliance, plug in what you need, skip what you don't. Self-host free.
Get Started Free Back to All Features

Related: deep-dives at Stripe and Shopify, comparisons against Synder and A2X, or the AI accounting story behind the GL postings.

---

## MCP Access: Use ERPClaw From Your AI Tools
URL: https://www.erpclaw.ai/features/mcp/

# Your business system, reachable from your AI tools
ERPClaw speaks MCP, the standard that lets AI assistants work with other software. Ask about your numbers from the assistant you already use, or wire ERPClaw into your own automations. Other products sell this as a feature; here it is simply how the system was built.
Developer docsGet ERPClaw free

## One action layer, every door

### Chat with it
The plain-language interface you use day to day. "Invoice Acme for the March retainer" is a complete instruction.

### Reach it from your assistant
MCP-capable AI tools connect to the same operations, so your business system joins the conversations you already have elsewhere.

### Program against it
Developers call the identical layer. No second-class API that drifts from what the product can really do.

Whichever door a request comes through, the same rules hold: deterministic accounting code does the posting and the checking engine verifies the books after every entry. Doors differ; the discipline does not.

## Common questions

### What is MCP, without the jargon?
MCP is the plug standard that lets AI assistants use other software. If a tool speaks MCP, an assistant like Claude can work with it directly: ask it questions, have it do things. ERPClaw speaks it natively.

### What can I actually do with it?
Work with your business system from the AI tools you already use. Ask about your numbers from your assistant, have it raise a document, pull a report into another workflow. The same plain-language operations, reachable from wherever you already work.

### Is this bolted on?
No, and that is the difference worth understanding. ERPClaw was built AI-native: every business operation was designed to be invoked by an AI in the first place. MCP access is that same layer exposed through the standard plug, not a wrapper added for marketing.

### Is it safe to let an AI touch my books?
The AI decides what you meant; deterministic accounting code decides what gets posted; and a checking engine verifies the books after every posting. That protection applies no matter which door the request came through, including MCP.

### What about developers?
Same surface, programmable: every business operation is callable, so integrations and automations use the exact layer the AI uses. The developer docs cover the details.

### Does it cost extra?
No. Like everything in ERPClaw, it is part of the free, self-hosted, open-source product.

Related: audit-ready books · what AI-native means · for developers

---

## Shopify Accounting Integration - Free A2X Alternative
URL: https://www.erpclaw.ai/features/shopify/

# The Shopify Integration That Paid Apps Charge $229/mo For.
Yours Free.
Covers every Shopify operation across 15 domains. Clearing-account reconciliation. Every order, fee, refund, payout, and gift card posted to your books, automatically. No middleware. No monthly fees. No lock-in.

Full
Coverage

Domains

Shared
Schema

$0
Cost / Year

## Why Not Just Use A2X?
A2X and similar paid Shopify accounting apps charge $29-229/month to sync your store to QuickBooks or Xero. Here is what you actually get for that money, and what you get for free.
| Feature | Paid Shopify Apps | ERPClaw
| Price | $29 to 229/month | $0 forever
| Middleware | Yes (syncs to QBO/Xero) | No (native ERP integration)
| COGS tracking | Premium tier only | Built-in
| Gift card accounting | Basic | Automatic deferred revenue
| Payout reconciliation | Summary journals | Transaction-level + clearing account
| Multi-location COGS | No | Yes
| Custom GL rules | No | Fully configurable
| Source code | Proprietary | Open-source

## How It Works
Other apps need you to connect Shopify to QuickBooks to your bank. ERPClaw needs one sentence.

Chat with ERPClaw
You:
I sell through Shopify

ERPClaw:
Got it. I'll set up Shopify integration. What's your store URL and API token?

You:
mystore.myshopify.com / shpat_abc123...

ERPClaw:
Connected. Store schema created. Syncing your Shopify data now.
1,247 orders synced
89 refunds synced
156 payouts synced
GL entries posted for all transactions
Shopify Clearing account balance: $0.00 (fully reconciled)

You:
Show me my COGS by product category

ERPClaw:
COGS breakdown (last 90 days):
Apparel: $18,400 | Accessories: $6,200 | Home: $3,100
Total COGS: $27,700 | Revenue: $89,300
Gross margin: 69.0%

The conversation above is an illustrative example. The order, refund, payout, and margin figures are sample data, not a live store sync.

## Key Features

### Auto-Sync
Orders, refunds, payouts, and adjustments sync automatically from Shopify. No manual downloads, no CSV imports, no third-party middleware sitting between your store and your books.

### Clearing Account Pattern
Every order posts to a Shopify Clearing account. When Shopify sends the payout, the clearing account zeros out. If it doesn't zero, something is wrong, and you'll know immediately.

### Automatic GL Posting
Every Shopify event creates GL entries. Orders post revenue and COGS. Fees post to expense. Refunds reverse. Gift cards create deferred revenue. No manual journal entries.

### COGS Tracking
Cost of goods sold calculated and posted automatically for every order line item. Multi-location inventory? ERPClaw tracks COGS per warehouse, not just a global average.

### Gift Card Accounting
Gift card sales post to Deferred Revenue (a liability). When redeemed, the liability reduces and revenue is recognized. Proper GAAP treatment, fully automated.

### Dispute Handling
Shopify chargebacks automatically create reserve entries. Won disputes reverse the reserve. Lost disputes write off the receivable. Full audit trail either way.

## GL Posting Rules
Most Shopify accounting apps hide their journal logic. We publish ours. Trust but verify.
| Event | Debit | Credit | Note
| Order | Shopify Clearing + Shopify Fees + COGS | Revenue + Inventory | Net to clearing, fees to expense, inventory to COGS
| Refund | Revenue + Inventory | Shopify Clearing + COGS | Reversal of original order entries
| Payout | Bank Account | Shopify Clearing | Clearing account should hit zero
| Fee | Shopify Fees (Expense) | Shopify Clearing | Transaction fees, subscription fees
| Gift Card Sale | Shopify Clearing | Deferred Revenue (Liability) | Revenue deferred until redemption
| Dispute | Dispute Reserve | Shopify Clearing | Reserve until resolved

## The Clearing Account Pattern
Shopify collects money from your customers, deducts fees, and deposits the rest into your bank 2-3 days later. The clearing account tracks every dollar in transit so nothing falls through the cracks.

### Order placed
Customer pays $100 on Shopify
DR Shopify Clearing $95 / CR Revenue $100, Fees $5 to Expense

### Shopify holds funds
Clearing account balance: $95

### Payout sent
Shopify deposits $95 to your bank
DR Bank Account $95 / CR Shopify Clearing $95

### Reconciled
Clearing account balance: $0
Clearing account = $0.00, fully reconciled

If the clearing account is not zero, it means Shopify owes you money, or there is a transaction that hasn't been matched. Either way, you'll know immediately, no month-end surprises.

## Every Industry, Not Just E-Commerce
Paid Shopify accounting apps only serve e-commerce stores. ERPClaw serves everyone who sells through Shopify.

### Medical Practice

Selling medical supplies through a Shopify storefront

### School

Selling uniforms, books, and supplies via Shopify

### Construction Company

Selling materials and equipment through Shopify

### Food & Beverage

D2C food sales with perishable inventory tracking

### Retail Brand

Multi-channel retail with Shopify as the primary storefront

### Nonprofit

Selling merchandise and accepting donations via Shopify

## The Open Source Advantage
Read our code. Run our tests. See exactly how every Shopify order becomes a GL entry.
Open
Every posting rule readable in the source

Tested
Automated tests run on every change

Six
Validation gates passed before release

### No Vendor Lock-in
If A2X raises prices or shuts down, your accounting integration disappears. With ERPClaw, you own everything: data, code, and configuration. Self-hosted on your machine, backed by an open-source community.

## Technical Specs
Tables
dedicated tables (shopify_order, shopify_payout, shopify_refund, etc.)

API Key
Private app credentials, only permissions ERPClaw needs, nothing more

Storage
API credentials encrypted at rest in ERPClaw's credential vault

SDK
shopify Python package auto-installed on first use

Webhook
Webhook endpoint auto-configured for real-time order and payout sync

Idempotency
Every Shopify event processed exactly once (order ID deduplication)

## Stop paying $229/month for something that should be free.
Install ERPClaw and say:
"I sell through Shopify"
Install in 5 Minutes View on GitHubCompare with A2X
Also using Stripe?See ERPClaw for Stripe

---

## Stripe Accounting Software - Free Rillet Alternative
URL: https://www.erpclaw.ai/features/stripe/

# Every Stripe Transaction Becomes a Journal Entry.
Automatically.
Individual transaction level, not daily summaries. Covers every Stripe operation across 10 domains. ASC 606 revenue recognition. 3-layer deterministic reconciliation. Setup takes under 5 minutes, no configuration screens. Rillet charges enterprise pricing. ERPClaw is free and open source.

Full
Coverage

Domains

Shared
Schema

$0
Cost / Year

## Ask it like you'd ask a person
"What did Stripe take from us last month?"
One ask. Your fees, refunds, and payout timing for any month, from your real books, not Stripe's dashboard math. And pretty much every Stripe question works the same way: payouts, refunds, disputes, subscription revenue, month by month or across any range you name.

## Tested Against Real Stripe Data
Not a demo with fake numbers. We connected a real Stripe sandbox and synced every transaction.
$8,474.51
Gross revenue synced

Charges processed

Payouts reconciled

Customers mapped

391 total records synced from a real Stripe sandbox, every one with its own GL entry.

## What Makes This Different
Replace hours of spreadsheet work with one chat command.

### Every Transaction Becomes a Journal Entry
Individual transaction level, not daily summaries. Every charge, refund, fee, and dispute creates its own GL entry with full Stripe metadata, the who, what, when, and how much. Zero manual data entry.

### 3-Layer Payout Reconciliation Without Spreadsheets
Layer 1: match Stripe transactions to payment entries. Layer 2: match payouts to bank deposits. Layer 3: verify the clearing account hits zero. One command replaces hours of spreadsheet work.

### Natural Language Accounting Through Chat
Say "sync my Stripe data" or "show me last month's MRR." Replace hours of spreadsheet work with one chat command. Setup takes under 5 minutes, no configuration screens, just paste your API key.

## Rillet Charges Enterprise Pricing. ERPClaw Is Free.
Companies like Rillet raised $100M+ to build Stripe-to-GL automation for SaaS companies. Here is what you actually get for enterprise pricing, and what you get for free.
| Feature | Paid Stripe ERPs | ERPClaw
| Price | Enterprise pricing (sales call required) | $0 forever, open source license
| Implementation | 45 days + CPA consulting team | Under 5 minutes, no configuration screens
| Data granularity | Individual transactions (SaaS only) | Individual transactions, every industry
| Reconciliation | ML-based (~95% auto-match) | Deterministic posting rules: the same input produces the same books, every time
| GL transparency | Proprietary rules, closed source | Published GL rules, see every debit/credit
| Industries | Venture-funded SaaS companies only | Healthcare, education, retail, construction, legal, 12+ more
| Data sovereignty | Their cloud (vendor lock-in) | Your machine, self-hosted, open-source
| Source code | Proprietary, closed | Open-source on GitHub
| Vendor risk | VC-funded startup burning cash | Open-source, community-owned, runs forever
| Stripe Connect | Basic payment sync | Application fees, transfers, connected account payouts
| Testing | Unknown (proprietary) | Automated regression suite plus a gated validation pipeline
| Platform | Standalone accounting tool | Built on OpenClaw, full ERP

## How It Works
Other ERPs need a 45-day implementation with a CPA consulting team. ERPClaw needs one conversation. This is a real tested flow, not a mockup.

Chat with ERPClaw on OpenClaw
You:
I just started a consulting firm in Austin, Texas. Help me get everything set up.

ERPClaw:
Done! Company created, consulting industry, Austin TX. Chart of accounts, fiscal year, and defaults configured.

You:
I process all my client payments through Stripe. Can you set that up?

ERPClaw:
Stripe integration ready. I can set up automated sync, invoice payment links, recurring billing, and automatic reconciliation. What's your restricted API key?

You:
rk_live_51abc...

ERPClaw:
Connected. Syncing your Stripe data now.
Charges synced with full metadata
Payouts reconciled to bank deposits
Customers mapped to your ERP
GL entries posted for every transaction

You:
Reconcile my Stripe payouts

ERPClaw:
Payout Reconciliation:
Gross Revenue: $8,474.51 | Refunds: ($248.47) | Net: $8,226.04
Stripe Fees: ($184.49) | To Bank: $7,950.62
14 payouts, all matched

You:
Show me my top Stripe customers

ERPClaw:
Top 5 by revenue:
1. Donald G., $1,400.64 (4 txns)
2. Brittany R., $1,019.89 (1 txn)
3. Taylor S., $929.86 (6 txns)
4. Richard T., $846.28 (19 txns)
5. Jeremy S., $622.16 (2 txns)
21 total customers | $8,474.51 total revenue

Based on a real end-to-end test: fresh install → company setup → Stripe connect → sync → reports. Verified March 2026.

## Key Features

### Auto-Sync
Charges, refunds, disputes, and payouts sync automatically. No manual downloads, no CSV imports, no reconciliation spreadsheets. Individual transaction level with full metadata, not daily summaries.

### 3-Layer Reconciliation
Layer 1: match Stripe transactions to ERPClaw payment entries. Layer 2: match payouts to bank deposits. Layer 3: verify GL clearing account hits zero. One command reconciliation.

### Automatic GL Posting
Every Stripe event creates GL entries. Charges post revenue. Fees post to expense. Refunds reverse. Disputes create reserves. Zero manual journal entries.

### Dispute Handling
Disputes automatically create reserve entries. Won disputes reverse the reserve. Lost disputes write off the receivable. Full audit trail either way.

### Stripe Connect
Multi-party payments for platforms and marketplaces. Application fees, transfers, and connected account payouts, all posted to the correct GL accounts automatically.

### MRR Reports
Monthly recurring revenue, churn rate, expansion revenue, and contraction, calculated from actual Stripe subscription data, not estimates.

## GL Posting Rules
Most ERPs hide their GL rules behind proprietary code. We publish ours. Every transaction becomes a journal entry you can trace and verify.
| Event | Debit | Credit | Note
| Charge | Stripe Clearing + Stripe Fees | Revenue | Net amount to clearing, gross fees to expense
| Refund | Revenue | Stripe Clearing | Reversal of original charge
| Dispute | Dispute Reserve | Stripe Clearing | Reserve until resolved
| Payout | Bank Account | Stripe Clearing | Clearing account should hit zero
| Fee Refund | Stripe Clearing | Stripe Fees | Stripe returns the fee on refund

## Every Industry, Not Just SaaS
Paid Stripe ERPs only serve venture-funded SaaS companies. ERPClaw serves everyone who accepts Stripe payments.

### Medical Practice

Collecting copays and patient payments via Stripe

### School

Collecting tuition and fees via Stripe

### Construction Company

Collecting progress payments via Stripe

### Law Firm

Collecting retainers and case payments via Stripe

### Retail Store

Processing card payments via Stripe Terminal

### Property Manager

Collecting rent and deposits via Stripe

## We Don't Guess. We Prove.
Deterministic reconciliation means every GL entry is traceable back to a Stripe balance_transaction. No AI guessing which invoice a payment belongs to.

### Their Approach

- ML-based matching algorithms
- ~95% auto-match rate
- 5% requires human intervention
- Can't explain why a match was made

### Our Approach

- Deterministic 3-layer matching
- balance_transaction → charge → payout
- Deterministic matching: the same input produces the same books
- Every GL entry traceable to source

Every GL entry references a Stripe balance_transaction. Every balance_transaction links to a charge or payout. There is no AI guessing which invoice a payment belongs to, Stripe tells us, and we post it.

## vs. Manual Reconciliation
Most businesses spend 3-5 days per month reconciling Stripe with their books. ERPClaw does it with one command, zero manual data entry.
| Aspect | ERPClaw + Stripe | Manual Process
| Reconciliation time | 0 days (automatic) | 3-5 business days
| Fee tracking | Per-transaction, automatic | Monthly aggregate
| Dispute visibility | Real-time GL reserve | Discovered at month-end
| Payout matching | Automatic to bank deposit | Manual spreadsheet
| Revenue recognition | Posted on charge | Batch at month-end
| Manual data entry | Zero | Every transaction
| Error rate | No manual keying: deterministic rules post the same books from the same input, every time | 2-5% industry average

## The Open Source Advantage
Read our code. Run our tests. See exactly how every Stripe charge becomes a GL entry. Rillet charges enterprise pricing. ERPClaw is free and open source.
Open
Every posting rule readable in the source

Tested
Automated tests run on every change

Six
Validation gates passed before release

### No Vendor Lock-in
If a paid Stripe ERP goes under, your accounting data is trapped in their cloud. With ERPClaw, you own everything, data, code, and configuration. Self-hosted on your machine, backed by an open-source community. Built on the OpenClaw platform.

## Technical Specs
Tables
dedicated tables (stripe_charge, stripe_payout, stripe_dispute, etc.)

API Key
Restricted key, only permissions ERPClaw needs, nothing more

Storage
API key encrypted at rest in ERPClaw's credential vault

SDK
stripe Python package auto-installed on first use

Webhook
Webhook endpoint auto-configured for real-time sync

Idempotency
Every Stripe event processed exactly once (idempotency key tracking)

Platform
Built on OpenClaw, the open-source AI agent platform

## Stop paying enterprise pricing for something that should be free.
Install ERPClaw on OpenClaw and say:
"I process payments through Stripe"
Install on Stripe Marketplace Try the Demo View on GitHubCompare with Rillet
Also using Shopify?See ERPClaw for Shopify

---

## Supply Chain ERP - Inventory + Manufacturing Free
URL: https://www.erpclaw.ai/features/supply-chain/

# Supply Chain management
Four tightly integrated skills covering inventory, selling, buying, and manufacturing. From raw materials to finished goods to customer delivery, managed through natural language.
Inventory (30)Selling (36)Buying (34)Manufacturing (24)
Inventory

Selling

Buying

## Inventory
Full coverage
Item master, warehouses, stock entries, batch and serial tracking, pricing rules, stock reconciliation, valuation, and CSV import.

### Items
`add-item`Create a new item in the catalog
`update-item`Modify item properties
`get-item`Retrieve item details
`list-items`Browse all items with filters
`list-item-groups`View item categories

### Warehouses
`add-warehouse`Register a new warehouse
`list-warehouses`Browse all warehouses
`get-warehouse-stock`Check stock at a warehouse

### Stock Entries
`add-stock-entry`Create a stock movement draft
`update-stock-entry`Modify a draft stock entry
`get-stock-entry`Retrieve stock entry details
`list-stock-entries`Browse all stock entries

### Operations
`submit-stock-entry`Finalize and post stock movement
`cancel-stock-entry`Reverse a posted stock entry
`create-stock-reconciliation`Start a stock count adjustment
`submit-stock-reconciliation`Post reconciliation adjustments
`get-stock-balance`Check current quantity on hand
`get-stock-valuation`Get current stock value
`list-stock-ledger-entries`View stock movement history
`validate-batch-expiry`Check batch expiration dates

### Pricing
`add-price-list`Create a new price list
`add-item-price`Set an item price
`add-pricing-rule`Create discount / pricing rule
`get-applicable-pricing-rule`Find matching pricing rules

### Batch / Serial
`add-batch`Register a new batch number
`add-serial-number`Register a serial number

### System
`status`Check skill health status
`csv-import`Import items from CSV file

### Example prompts
"Add item Widget A at $25""Check stock balance""Transfer 50 units to Warehouse B""Import items from CSV"

## Selling
Full coverage
Full order-to-cash cycle: customers, quotations, sales orders, delivery notes, sales invoices, credit notes, recurring invoices, and sales partners.

### Customers
`add-customer`Create a new customer record
`update-customer`Modify customer details
`get-customer`Retrieve customer information
`list-customers`Browse all customers

### Quotations
`add-quotation`Create a sales quotation draft
`update-quotation`Modify a draft quotation
`get-quotation`Retrieve quotation details
`list-quotations`Browse all quotations
`submit-quotation`Finalize a quotation
`convert-quotation-to-so`Convert quotation to sales order

### Sales Orders
`add-sales-order`Create a sales order draft
`update-sales-order`Modify a draft sales order
`get-sales-order`Retrieve sales order details
`list-sales-orders`Browse all sales orders
`submit-sales-order`Finalize and confirm sales order
`cancel-sales-order`Cancel a submitted sales order

### Delivery Notes
`create-delivery-note`Create DN from sales order
`get-delivery-note`Retrieve delivery note details
`list-delivery-notes`Browse all delivery notes
`submit-delivery-note`Post delivery and update stock
`cancel-delivery-note`Reverse a posted delivery

### Sales Invoices
`create-sales-invoice`Create invoice from SO or DN
`update-sales-invoice`Modify a draft invoice
`get-sales-invoice`Retrieve invoice details
`list-sales-invoices`Browse all sales invoices
`submit-sales-invoice`Post invoice and GL entries
`cancel-sales-invoice`Reverse a posted invoice

### Credit Notes
`create-credit-note`Create credit note from invoice

### Partners
`add-sales-partner`Register a sales partner
`list-sales-partners`Browse all sales partners

### Recurring
`add-recurring-template`Create a recurring invoice template
`update-recurring-template`Modify a recurring template
`list-recurring-templates`Browse recurring templates
`generate-recurring-invoices`Generate invoices from templates

### Other
`update-invoice-outstanding`Recalculate outstanding amount
`status`Check skill health status

### Example prompts
"Create a sales order for 50 widgets""Invoice Acme Corp $5,000""Create delivery note from SO-2026-00001"

### Order to Cash Workflow
Quotation Sales Order Delivery Note Sales Invoice Payment
Each step creates the next document automatically. GL entries and stock ledger update at every posting point.

## Buying
Full coverage
Procure-to-pay: suppliers, material requests, RFQs, supplier quotations, purchase orders, purchase receipts, and purchase invoices.

### Suppliers
`add-supplier`Create a new supplier record
`update-supplier`Modify supplier details
`get-supplier`Retrieve supplier information
`list-suppliers`Browse all suppliers

### Material Requests
`add-material-request`Create a material request
`update-material-request`Modify a material request
`get-material-request`Retrieve request details
`list-material-requests`Browse all material requests

### RFQ
`add-rfq`Create a request for quotation
`update-rfq`Modify an RFQ draft
`get-rfq`Retrieve RFQ details
`list-rfqs`Browse all RFQs

### Supplier Quotes
`add-supplier-quotation`Record a supplier quote
`list-supplier-quotations`Browse supplier quotes

### Purchase Orders
`add-purchase-order`Create a purchase order draft
`update-purchase-order`Modify a draft PO
`get-purchase-order`Retrieve PO details
`list-purchase-orders`Browse all purchase orders
`submit-purchase-order`Finalize and confirm PO
`cancel-purchase-order`Cancel a submitted PO

### Purchase Receipts
`create-purchase-receipt`Create receipt from PO
`get-purchase-receipt`Retrieve receipt details
`list-purchase-receipts`Browse all purchase receipts
`submit-purchase-receipt`Post receipt and update stock
`cancel-purchase-receipt`Reverse a posted receipt

### Purchase Invoices
`create-purchase-invoice`Create invoice from PO or receipt
`update-purchase-invoice`Modify a draft purchase invoice
`get-purchase-invoice`Retrieve purchase invoice details
`list-purchase-invoices`Browse all purchase invoices
`submit-purchase-invoice`Post invoice and GL entries
`cancel-purchase-invoice`Reverse a posted purchase invoice

### System
`status`Check skill health status

### Example prompts
"Add supplier Widget Corp""Create PO for 200 widgets at $18""Receive goods from PO-2026-00005"

## Manufacturing
Full coverage
Bill of materials, work orders, job cards, production planning (MRP), operation tracking, subcontracting, and product costing.

### BOM
`add-bom`Create a bill of materials
`add-bom-item`Add component to a BOM
`add-bom-operation`Add operation step to a BOM
`get-bom`Retrieve BOM details
`list-boms`Browse all BOMs
`update-bom`Modify a BOM

### Operations
`add-operation`Define a manufacturing operation
`add-workstation`Register a workstation
`add-routing`Create a manufacturing routing
`list-routings`Browse all routings

### Work Orders
`add-work-order`Create a work order
`get-work-order`Retrieve work order details
`list-work-orders`Browse all work orders
`submit-work-order`Start production on a work order
`close-work-order`Complete a work order

### Production Planning
`add-production-plan`Create a production plan
`add-production-plan-item`Add item to production plan
`list-production-plans`Browse all production plans
`suggest-material-request`Auto-suggest materials needed

### Job Cards
`add-job-card`Create a job card for an operation
`add-subcontracting-order`Create a subcontracting order
`submit-subcontracting-order`Confirm subcontracting order

### System
`calculate-product-costing`Calculate total product cost
`status`Check skill health status

### Example prompts
"Create BOM for Widget A""Start work order for 50 units""Calculate product cost"

## Run your supply chain from chat
full coverage across inventory, selling, buying, and manufacturing. All included free.
Get Started Free

Related: read the architecture in AI-native ERP, see the foundation features, or compare against the entrenched options at NetSuite and Odoo. Inventory deep-dive: FIFO vs weighted average.

---

## Three-Way Matching: Check Every Bill Before It's Paid
URL: https://www.erpclaw.ai/features/three-way-matching/

# The bill, the order, and the delivery have to agree
Three-way matching is the oldest trick in good bookkeeping: never pay a bill until it matches what you ordered and what actually showed up. ERPClaw does it on every purchase, automatically, so the expensive little mistakes get caught while they are still cheap.
How purchasing worksGet ERPClaw free

## Three documents, one truth

### 1. What you ordered
"Order 100 steel brackets from Acme at 10 dollars each." The purchase order records the deal as agreed.

### 2. What arrived
"The Acme delivery is here, 95 brackets, 5 short." The receipt records reality at the dock, not the paperwork's version of it.

### 3. What they billed
The supplier's invoice lands and is checked against both. Ninety-five brackets arrived; a bill for 100 does not get paid quietly.

When all three agree, payment proceeds and the books record the whole chain. When they do not, a person decides, and the decision is on the record. That is the difference between a control and a habit.

## Common questions

### What is three-way matching, in plain terms?
Before a supplier bill gets paid, three documents have to agree: what you ordered, what actually arrived, and what the supplier billed. If the bill says 120 units at 11 dollars and the order said 100 at 10, that gap is caught before the payment, not after.

### What kinds of mistakes does it catch?
The classics: paying for goods that never arrived, quantities that grew between order and invoice, prices that drifted from what was agreed, and the same bill showing up twice. Individually small, together they are real money every year.

### Do I have to set it up?
No. It is how ERPClaw's purchasing flow works: order, receipt, bill, each recorded as its own step and checked against the others. You just run purchasing normally, in plain language.

### What happens when something does not match?
The mismatch is surfaced for a human decision instead of sliding through. Sometimes the price really did change; the point is that someone chose to accept it, and that choice is on the record.

### Is this a paid add-on?
No. ERPClaw is free forever, self-hosted, open source, and matching is part of the purchasing flow, not a tier.

Related: audit-ready books · bank statement matching · the supply chain features

---

## ERP for Accountants - Audit-Ready GL, Double-Entry Enforced
URL: https://www.erpclaw.ai/for/accountants/

# Built for
Accounting Precision
Immutable general ledger with SHA-256 chain hashing. Double-entry enforced by 12-step validation, with invariant checks on every batch. All testing gates passed, GL balanced to the cent. This is accounting software that respects the profession.

## Why accountants love ERPClaw
No shortcuts. No workarounds. Every design decision prioritizes data integrity and auditability.
Immutable General LedgerEvery GL entry is protected by SHA-256 chain hashing. No entry can be silently altered. Corrections are made through proper reversals, creating a complete, tamper-evident audit trail.

12-Step GL ValidationEvery posting passes the full set of mandatory checks: balanced debits and credits, valid fiscal period, account type validation, cost center assignment, currency consistency, and more. No exceptions, no overrides.

Instant Financial ReportsTrial balance, profit and loss, balance sheet, cash flow statement, general ledger detail, AR/AP aging, all generated on demand from a single chat command. No waiting, no report builder.

Multi-Company ConsolidationManage multiple entities from one system with inter-company journal entries and consolidated reporting. Each company maintains its own chart of accounts, fiscal years, and cost centers.

## Compliance built in, not bolted on
US-focused compliance features designed for how businesses actually operate.

### 1099 Reporting
Track vendor payments and generate 1099-NEC/MISC data with automatic threshold monitoring

### Tax Templates
US sales tax templates with configurable rates, tax categories, and multi-jurisdiction support

### Wage Garnishment
Payroll deduction tracking for garnishments with priority ordering and limit enforcement

### Budget Controls
Budget vs actuals at GL account level with configurable overspend alerts and period controls

### Intercompany Elimination
Automatic elimination entries for inter-company transactions during consolidated reporting

### GL Integrity Checksums
Database-level integrity verification ensures no entry has been altered outside the application

## 20 reports, one chat command each
Every report is generated instantly from your live data. No report builder, no export steps. Just ask.
| # | Report | Module | Command
| 1 | Trial Balance | Reports | `get-trial-balance`
| 2 | Profit & Loss | Reports | `get-profit-and-loss`
| 3 | Balance Sheet | Reports | `get-balance-sheet`
| 4 | Cash Flow Statement | Reports | `get-cash-flow`
| 5 | General Ledger Detail | Reports | `get-general-ledger`
| 6 | Accounts Receivable Aging | Reports | `get-accounts-receivable`
| 7 | Accounts Payable Aging | Reports | `get-accounts-payable`
| 8 | Budget vs Actuals | Reports | `get-budget-variance`
| 9 | Gross Profit Report | Reports | `get-gross-profit`
| 10 | Bank Reconciliation | Payments | `get-bank-reconciliation`
| 11 | Sales Register | Reports | `get-sales-register`
| 12 | Purchase Register | Reports | `get-purchase-register`
| 13 | Stock Ledger Report | Inventory | `get-stock-ledger-report`
| 14 | Item-wise Profitability | Analytics | `get-item-profitability`
| 15 | Revenue Trend Analysis | Analytics | `get-revenue-trend`
| 16 | Expense Trend Analysis | Analytics | `get-expense-trend`
| 17 | Financial Ratios | Analytics | `get-financial-ratios`
| 18 | Payroll Summary | Payroll | `get-payroll-summary`
| 19 | Project Profitability | Projects | `get-project-profitability`
| 20 | Cash Flow Forecast | AI Engine | `get-cash-flow-forecast`

## See it for yourself
Run a trial balance in under a minute. Free, self-hosted, no strings attached.
Start with a Trial Balance See All Features

---

## ERPClaw for livestock: AI-native, USDA-ready, free
URL: https://www.erpclaw.ai/for/agriculture/livestock/

# ERPClaw for livestock operations
Herd records, breeding, feed inventory, vaccination compliance, USDA reporting, payroll. AI-native, $0 forever.
36 livestock-specific actions. 15 livestock-specific tables.

## What ERPClaw does for livestock operations
Livestock operation management with herd records, breeding tracking, feed inventory, vaccination logs, weight tracking, USDA reporting, and full back-office ERP. Built for cattle, swine, dairy, sheep, and mixed operations.

## Capabilities

### Herd records + animal-level traceability
Per-animal records (ear-tag, RFID, or pen-level) with breed, birthdate, sire/dam lineage, weight history, vaccinations, treatments. The shared database means a feeder steer's full history follows it through every operation.

### Breeding + calving records
Breeding events, expected due dates, calving outcomes, weaning weights. Auto-generated reminders for pregnancy checks, vaccination windows, and weaning dates.

### Feed inventory with cost-per-pound
Bulk feed inventory with FIFO costing, ration formulation, and feed-conversion-ratio tracking per pen or per animal. The AI agent computes feed cost per pound of gain.

### Vaccination + treatment compliance
Per-animal vaccination logs, treatment records with withdrawal-time tracking, and herd-level compliance reporting. Withdrawal-time alerts before sale or harvest.

### USDA + state reporting
Common USDA report formats (mandatory inventory reports for certain operations, premise ID tracking, brand-inspection logs in brand states). Self-hosted means your operation data stays on your hardware.

## Frequently asked questions

### How does ERPClaw compare to CattleMax or Performance Beef?
CattleMax and Performance Beef are mature livestock-specific platforms with deeper field-data-collection and equipment integrations. ERPClaw covers herd records, breeding, vaccinations, feed inventory, USDA reporting, and the full back-office ERP (accounting, payroll, equipment depreciation). For operations that want one system end-to-end with AI-native architecture and $0 forever cost, ERPClaw is the AI-native option.

### Does ERPClaw support RFID + EID tag readers?
ERPClaw stores RFID and visual tag IDs and supports CSV import from common tag readers. Direct hardware integration is added based on operation demand. For a 500-head feedlot with weekly weigh-ins, the CSV-import workflow is fast enough.

### What about commodity hedging and grain marketing?
ERPClaw's foundation accounting handles forward contract recording, commodity inventory, and basis tracking. Deep grain-marketing analytics (DTN-style decision support) are not in scope today; we recommend pairing with a dedicated grain-marketing tool if that is a primary need.

### Is the agriculture module free too?
Yes. agricultureclaw is part of ERPClaw's open-source build. All modules including the 36 livestock-specific actions are Free &amp; Open Source and $0 forever. Self-hosted on your hardware or a small VPS.

Related: read the AI ERP buyer's guide, the AI inventory guide, the test surface, or the install guide.

---

## Residential Construction ERP: Free, AI-Native
URL: https://www.erpclaw.ai/for/construction/residential/

# ERPClaw for residential construction
Job costing, change orders, AIA progress billing, lien waivers, subcontractor management, payroll. AI-native, $0 forever.
47 construction-specific actions. 18 construction-specific tables.

## What ERPClaw does for residential GCs
Residential construction project management with job costing, change orders, subcontractor management, progress billing, lien-waiver tracking, and full back-office ERP. Built specifically for residential GCs and remodelers.

## Capabilities

### Job costing with phase + cost-code structure
Track costs by job, phase (excavation, framing, MEP rough, finish), and CSI cost code. Variance against estimate is real-time, not month-end. The AI agent flags overruns when they happen, not after the close.

### Change orders + customer approval flow
Create change orders with line-item pricing, route for customer approval (digital or printed), update the job cost baseline on approval. The audit trail preserves every revision; nothing is silently overwritten.

### Subcontractor management + 1099 tracking
Vendor records with W-9 capture, certificate of insurance tracking, retention holdback, and 1099-NEC year-end reporting integrated with the foundation finance module.

### Progress billing (AIA-style G702/G703)
Application-for-payment generation with completed-to-date and retainage columns. Tied to the GL through the foundation accounting module so revenue recognition follows percentage-of-completion automatically.

### Lien waiver tracking
Conditional vs unconditional, partial vs final lien waivers tracked per subcontractor per job. Required-but-missing waivers surface before you cut the check, not after.

## Frequently asked questions

### How does ERPClaw compare to Buildertrend or CoConstruct?
Buildertrend and CoConstruct are mature residential-construction project management platforms with deeper customer-portal and selections-management features. ERPClaw covers job costing, change orders, subcontractor management, progress billing, lien waivers, and the full back-office ERP (accounting, payroll, inventory). For builders who want one system end-to-end with AI-native architecture and $0 forever cost, ERPClaw is the AI-native option.

### Does ERPClaw handle AIA G702/G703 progress billing?
Yes. The Projects + Billing modules support AIA-style applications for payment with the standard G702 cover sheet and G703 schedule of values. Tied to the GL so percentage-of-completion revenue recognition runs automatically.

### What about integration with QuickBooks for accounting?
ERPClaw replaces QuickBooks entirely; you do not need to integrate. The accounting module is included in the open-source build with immutable GL, US tax forms, payroll, and 20-plus financial reports. The /migrate/from-quickbooks/ playbook walks through the move.

### Is ERPClaw too much ERP for a 5-person residential GC?
No. You install the foundation plus the constructclaw module and ignore what you don't need. A 5-person GC uses job costing, change orders, AP for subs, and payroll. The rest of the catalog is there if you grow into it but they don't get in the way.

Related: read the AI ERP buyer's guide, the migration from QuickBooks, the test surface, or the install guide.

---

## School ERP Buyer Guide for K-12 and Higher Education
URL: https://www.erpclaw.ai/for/education/

# A practical ERP guide for
schools
every operation. K-12 and higher ed. Students, grades, attendance, state reporting.

Full
Coverage

Whole ERP
Scope

FERPA
Compliance

Ed-Fi
Integration

## Every module, Every Level
From elementary schools to universities. ERPClaw detects your institution type and installs the right modules automatically.

### EduClaw Core
EduClaw Core
Students, enrollment, attendance, grades, academic years, programs, courses, sections, instructors, rooms, and academic calendar management.

### EduClaw K-12
EduClaw K-12
Grade levels, homerooms, behavior tracking, parent contacts, report cards, grade-level standards, and elementary/secondary specific workflows.

### EduClaw Financial Aid
EduClaw Financial Aid
FAFSA processing, aid packaging, scholarship management, disbursement scheduling, satisfactory academic progress, and R2T4 calculations.

### EduClaw Scheduling
EduClaw Scheduling
Timetable auto-generation, room conflict detection, instructor availability, section capacity management, and waitlist processing.

### EduClaw LMS
EduClaw LMS
Assignments, submissions, grading rubrics, discussion forums, content modules, learning outcomes, and gradebook synchronization.

### EduClaw State Reporting
EduClaw State Reporting
Ed-Fi data standard compliance, state enrollment reporting, demographic submissions, and automated validation against state rules.

### EduClaw Higher Ed
EduClaw Higher Ed
Degree audit, transfer credits, GPA calculation, academic standing, commencement, transcript generation, and registrar workflows.

New

## Parent Portal
12 FERPA-compliant actions. Parents see grades, attendance, schedules, and meal balances, nothing more. Every access logged.

parent-view-grades
View child's current grades and GPA

parent-view-attendance
See attendance record and absences

parent-view-schedule
View class schedule and room assignments

parent-view-assignments
See upcoming and past assignments

parent-message-teacher
Send FERPA-compliant messages to teachers

parent-view-report-card
Download official report cards

parent-update-contact
Update emergency contact information

parent-view-fees
View and pay outstanding balances

parent-excuse-absence
Submit absence excuses with documentation

parent-opt-out
Manage directory information opt-out per FERPA

parent-view-bus
View bus route and stop assignment

parent-view-meal-balance
Check cafeteria account balance

FERPA-compliant, directory opt-out, access logging, role-based visibility

New

## Cafeteria & NSLP
USDA National School Lunch Program compliance. Track meals, calculate reimbursements, and manage student meal accounts.

Meal Tracking
Track free, reduced, and paid meals per student per day

NSLP Compliance
National School Lunch Program reporting and reimbursement claims

USDA Reimbursement
Auto-calculate federal reimbursement rates by meal category

Meal Patterns
Verify meals meet USDA nutritional requirements

Account Management
Prepaid meal accounts, low balance alerts, auto-replenishment

Eligibility
Free/reduced meal eligibility determination and verification

New

## Transportation
Bus routes, stops, student assignments, driver credentials, and ridership reporting for state transportation funding.

Route Management
Define bus routes with ordered stops, timing, and capacity

Stop Assignment
Assign students to stops based on address and eligibility

Driver Tracking
Track driver certification, CDL expiration, and background checks

Vehicle Maintenance
Bus inspection schedules, repair history, and mileage tracking

Ridership Reports
Daily ridership counts for state transportation funding

Hazard Zones
Manage walk zones and hazardous route exceptions

## And everything else

Professional Development
PD hours tracking, certification requirements, and workshop registration for faculty

Library Management
Catalog, circulation, overdue tracking, and patron management integrated with student records

Student Housing
Room assignments, occupancy management, meal plan enrollment, and housing contracts for higher ed

Activities & Clubs
Extracurricular management, eligibility verification, roster tracking, and activity fee collection

Timetable Auto-Generation
Constraint-based scheduling engine that handles room conflicts, instructor availability, and section balancing

Discipline & Behavior
Incident tracking, behavior interventions, referral management, and restorative justice workflows

## Setup in one conversation

Chat with ERPClaw
You:
I run a school

ERPClaw:
Setting up EduClaw. What kind of school?

You:
K-8, about 400 students, we need attendance, grades, and state reporting

ERPClaw:
Done. Here's what I set up:
, EduClaw Core installed (every operation)
, EduClaw K-12 installed (grade levels K through 8)
, EduClaw State Reporting installed (Ed-Fi ready)
, Academic year 2025-26 created with 2 semesters
, Parent portal enabled (12 FERPA-compliant actions)
, Cafeteria and transportation modules ready
Try: "Enroll a new student" or "Set up my class sections"

## Start managing your school
Install ERPClaw and say:
"I run a school"
ERPClaw handles the rest.
Get Started Free View on GitHub

Related: read the architecture in AI-native ERP, the books-side detail in AI accounting, or honest comparisons against NetSuite and ERPNext for school finance + student records. Get install steps in core docs.

---

## Healthcare ERP Software - HIPAA-Compliant, Free, AI-Native
URL: https://www.erpclaw.ai/for/healthcare/

# AI-Native
Healthcare ERP
every operation. Patients, billing, RCM, HIPAA. Just talk to it.
By Nikhil Jathar, Co-founder, ERPClaw · Published 2026-03-23 · Reviewed 2026-07-01

Full
Coverage

Shared
Schema

Whole ERP
Scope

Domains

## Every module, One Platform
Core clinical workflows plus 4 specialized sub-verticals. Say what kind of practice you run and ERPClaw installs the right modules automatically.

### HealthClaw Core
HealthClaw Core
Patients, encounters, vitals, diagnoses, prescriptions, lab orders, immunizations, provider management, care teams, and clinical workflows across 11 domains.

### HealthClaw Dental
HealthClaw Dental
Dental charting, tooth-level tracking, periodontal exams, treatment plans, CDT procedure codes, and dental-specific insurance billing.
More on ERPClaw for dental practices

### HealthClaw Vet
HealthClaw Vet
Species and breed tracking, vaccination schedules, weight history, boarding management, and multi-animal household records.
More on ERPClaw for veterinary clinics

### HealthClaw Mental
HealthClaw Mental
Session notes, treatment goals, PHQ-9/GAD-7 assessments, therapy modalities, group sessions, and outcomes tracking.

### HealthClaw Home Health
HealthClaw Home Health
Visit scheduling, OASIS assessments, caregiver assignment, mileage tracking, home safety evaluations, and Medicare compliance.

## Revenue Cycle Management
From eligibility check to payment posting. The entire revenue cycle managed through conversation.
Step 1
Payer Registry
Maintain payer contracts, fee schedules, and enrollment status

Step 2
Eligibility Check
Verify patient insurance before the visit

Step 3
Claim Scrubbing
Validate CPT/ICD codes, modifiers, and bundling rules before submission

Step 4
Claim Submission
Generate and submit CMS-1500 / 837P claims electronically

Step 5
ERA/835 Processing
Auto-post remittance advice and match to claims

Step 6
Denial Management
Track denials by reason code, auto-queue for resubmission

All steps automated. GL entries posted at each stage.

## Compliance Built In
HIPAA, No Surprises Act, MIPS. Not bolted on as afterthoughts. Built into the data model from day one.

### HIPAA PHI Audit
Every access to protected health information is logged. View who accessed what patient record, when, and why. Immutable audit trail.

### Good Faith Estimate
Auto-generate Good Faith Estimates for uninsured/self-pay patients as required by the No Surprises Act. CPT-based cost calculation.

### MIPS Quality Measures
Track Merit-based Incentive Payment System quality measures. Automated measure calculation and reporting period management.

### Business Associate Agreements
Track BAA status for all vendors and partners who handle PHI. Expiration alerts and renewal workflows.

### Breach Tracking
Log and manage potential PHI breaches. Risk assessment scoring, notification tracking, and HHS reporting workflows.

### Provider Credentialing
Track license expiration, board certification, DEA registration, malpractice insurance, and hospital privileges for every provider.

## Clinical Workflows

Immunizations
CDC schedule tracking, vaccine inventory, lot number recording, and immunization registry reporting

Medication Reconciliation
Compare patient-reported medications against prescribed, flag interactions and duplicates

Provider Credentials
License, DEA, NPI, board certification, and malpractice insurance with expiration alerts

Care Team Management
Assign primary, specialist, and allied health providers. Role-based access to patient records

Patient Merge
Safely merge duplicate patient records. All encounters, billing, and history consolidated with full audit trail

Lab Orders & Results
Order labs, track specimen collection, receive results, and flag abnormal values automatically

## Setup in one conversation

Chat with ERPClaw
You:
I run a medical practice

ERPClaw:
Setting up HealthClaw. What kind of practice?

You:
Family medicine, 3 providers, we take insurance

ERPClaw:
Done. Here's what I set up:
&bull; HealthClaw Core installed (140+ clinical actions)
&bull; 3 provider records created
&bull; Payer registry initialized
&bull; HIPAA audit logging enabled
&bull; CMS-1500 claim templates ready
Try: "Register a new patient" or "Show me today's schedule"

## What healthcare organizations evaluate when choosing an ERP
Feature lists look similar across healthcare ERP options. The four criteria below are where implementations actually differ, and where the open-source self-hosted model changes the calculus.

### Compliance audit trails
HIPAA requires that every access to protected health information be logged with who accessed it, when, and for what purpose. ERPClaw writes an immutable, chain-hashed audit log for every patient record read or write. The HHS Office for Civil Rights publishes the full technical safeguard requirements at hhs.gov/hipaa; ERPClaw's data model is built around those controls, not retrofitted to pass a checklist.
HHS HIPAA technical safeguards

### EHR integration readiness
Healthcare organizations ask whether a new ERP can exchange data with their existing electronic health record. ERPClaw exposes HL7 FHIR-compatible endpoints for patient demographics, encounter summaries, and lab results. This means scheduling, billing, and clinical data stay in sync without manual re-entry between systems.

### Multi-site billing
Group practices, health systems, and FQHCs run billing across multiple tax identification numbers, NPI numbers, and service locations. ERPClaw's billing engine supports separate payer contracts, fee schedules, and claim submission queues per location, with consolidated GL reporting rolled up to the parent organization.

### ASC 606 and grant revenue
Federally Qualified Health Centers, rural health clinics, and nonprofit hospital systems receive a mix of grant funding, capitation payments, and fee-for-service revenue. ERPClaw handles ASC 606 revenue recognition across all three streams: performance-obligation tracking for grants, period allocation for capitation, and claim-basis recognition for FFS. Donation revenue follows the same framework.

CMS publishes the current MIPS quality measure set and reporting requirements at qpp.cms.gov.

## Frequently asked questions

### Is ERPClaw free for healthcare organizations?
Yes. ERPClaw is open source under the Free &amp; Open Source license and costs $0 forever. You self-host it on your own hardware or cloud infrastructure. There is no per-seat pricing, no subscription tier, and no vendor lock-in. The full HealthClaw suite including all five modules and 230+ clinical actions is included.

### Is ERPClaw HIPAA-compliant ERP software?
ERPClaw provides the technical controls that HIPAA's Security Rule requires: immutable PHI audit logs, role-based access controls, encrypted credential storage, and a self-hosted deployment model that keeps patient data on your infrastructure. HIPAA compliance is a program, not a product certification. Your organization still needs a Business Associate Agreement, workforce training, and policies. ERPClaw handles the technical safeguard layer; the administrative and physical safeguards are your responsibility. See the HHS HIPAA technical safeguard guidance for the full requirement set.

### How does ERPClaw handle healthcare revenue cycle management?
ERPClaw includes a six-step RCM workflow: payer registry, eligibility verification, claim scrubbing (CPT/ICD validation), CMS-1500 and 837P claim submission, ERA/835 auto-posting, and denial management with reason-code tracking. Each step posts the corresponding GL entry automatically. This is not a third-party integration; the RCM engine is part of the open-source build.

### Can ERPClaw be used as an ERP for a small medical practice?
Yes. ERPClaw is designed to be installed and configured through a single conversation. For a small family medicine or specialty practice, you say what kind of practice you run and ERPClaw sets up patient records, scheduling, payer registry, and HIPAA audit logging in that session. There is no professional services engagement required to get started. Larger multi-provider practices benefit from the same platform with additional configuration for each provider's NPI and payer contracts.

### Does ERPClaw integrate with existing EHR systems?
ERPClaw exposes HL7 FHIR-compatible endpoints for patient demographics, encounter summaries, and lab data. Practices that want to run ERPClaw alongside an existing clinical EHR can exchange patient and billing data through these endpoints. ERPClaw is designed to own the back-office and RCM layer; it can complement a clinical EHR rather than replace it if that is the preference.

### Does ERPClaw support ASC 606 revenue recognition for healthcare grants?
Yes. FQHCs, rural health clinics, and nonprofit health systems receive grant funding, capitation payments, and fee-for-service revenue under different recognition rules. ERPClaw tracks performance obligations for grants, allocates capitation over the service period, and recognizes FFS revenue on a claim basis. Donation revenue follows the same ASC 606 framework with donor-restriction tracking.

### What does MIPS quality reporting look like in ERPClaw?
ERPClaw tracks Merit-based Incentive Payment System quality measures as a built-in compliance feature, not a paid add-on. Measures are calculated automatically from clinical data as encounters are recorded. The CMS Quality Payment Program defines the current measure set; ERPClaw maps clinical actions to those measures and produces reporting-period summaries.

### Can ERPClaw handle billing for multiple clinic locations?
Yes. Multi-site healthcare organizations can configure separate NPI numbers, tax identification numbers, payer contracts, and fee schedules per location. Claims submit under the correct billing entity for each site, and the GL consolidates all locations into a single chart of accounts for financial reporting at the parent organization level.

## Start managing your practice
Install ERPClaw and say:
"I run a medical practice"
That's the entire setup.
Get Started Free View on GitHub

Related: read the architecture in AI-native ERP, the books-side detail in AI accounting, or honest comparisons against NetSuite and Sage Intacct for healthcare practices. Get install steps in core docs.

---

## ERPClaw for dental practices: AI-native, HIPAA-aligned, free
URL: https://www.erpclaw.ai/for/healthcare/dental/

# ERPClaw for dental practices
Patient records, treatment plans, insurance claims, production reporting, payroll. AI-native, HIPAA-aligned, $0 forever.
38 dental-specific actions. 14 dental-specific tables. Built on the healthclaw foundation.

## What ERPClaw does for dental practices
Dental-specific patient records, treatment plans, insurance claims, recall scheduling, perio charting integrations, and HIPAA-aligned audit trail. Built on the healthclaw foundation with dental-specific extensions.

## Capabilities

### Patient records + treatment plans
Dental-specific patient records with medical history, allergies, current medications, and treatment plans. Multi-provider scheduling for dentists, hygienists, and assistants on shared calendars.

### Insurance claims + benefits verification
Pre-treatment estimates, claim submission, EOB matching, and patient-portion calculation. The AI agent handles the routine claims; ambiguous denials surface for human review.

### Recall + appointment reminders
Automated 6-month recall scheduling, treatment-plan-followup, and missed-appointment workflows. Patient communication templates for SMS or email.

### Production reporting + provider compensation
Per-provider production reports, hygienist productivity, treatment-plan acceptance rates, and compensation calculations integrated with the foundation payroll module.

### HIPAA-aligned audit trail
Immutable GL plus chain-hashed audit trail for every patient record access. Self-hosted deployment means PHI stays on your hardware, simplifying HIPAA compliance perimeter.

## Frequently asked questions

### Is ERPClaw HIPAA-compliant for dental practices?
ERPClaw provides the technical controls HIPAA requires (immutable audit trail, encrypted credentials, role-based access, self-hosted data residency). HIPAA compliance is ultimately a function of your business associate agreement, training, and policies; the software meets the technical requirements. Self-hosting on your hardware simplifies the compliance perimeter compared to cloud SaaS.

### How does ERPClaw compare to Dentrix or Eaglesoft?
Dentrix and Eaglesoft are mature practice-management systems with deeper dental-specific workflows (perio charting UI, X-ray imaging integration). ERPClaw covers patient records, scheduling, insurance claims, production reporting, and the full back-office ERP (accounting, HR, payroll). For practices that want one system end-to-end with AI-native architecture and $0 forever cost, ERPClaw is the AI-native option. For deep imaging-integrated workflows, Dentrix still leads on the operatory side.

### What about insurance integration with major carriers?
ERPClaw supports the standard 837/835 claim format. Carrier-specific eligibility-verification integrations are added based on practice demand. Talk to us if you need a specific carrier prioritized.

### Is the dental module free too?
Yes. healthclaw-dental is part of ERPClaw's open-source build. Every module including the 38 dental-specific actions are Free &amp; Open Source and $0 forever. Self-hosted on your hardware.

Related: read the broader healthcare page, the veterinary version, the AI ERP buyer's guide, the test surface, or the install guide.

---

## ERPClaw for veterinary clinics: AI-native, multi-pet, free
URL: https://www.erpclaw.ai/for/healthcare/veterinary/

# ERPClaw for veterinary clinics
Multi-pet records, vaccine reminders, lab integrations, treatment plans, inventory, payroll. AI-native, $0 forever.
32 vet-specific actions. 12 vet-specific tables. Built on the healthclaw foundation.

## What ERPClaw does for veterinary clinics
Veterinary patient records, multi-pet client management, vaccine reminders, lab integrations, breed-specific protocols, and full back-office ERP. Built on the healthclaw foundation with vet-specific extensions.

## Capabilities

### Multi-pet client records
One client record links to multiple patients (pets). Per-patient medical history, vaccinations, allergies, weight tracking, and breed-specific protocol templates. Multi-doctor scheduling on shared calendars.

### Vaccination + recall reminders
Per-patient vaccine schedules with auto-reminders by SMS or email. Annual exam recall, dental cleaning, parasite-prevention reminders. Compliance reporting for state-required rabies records.

### Treatment + lab integrations
Treatment plan templates for common procedures. Lab result imports from common veterinary diagnostic services. Imaging integration ready for typical clinic equipment.

### Inventory for medications + supplies
Stock control for prescription medications, vaccines, surgical supplies. FIFO costing, multi-location for clinics with multiple sites, expiration date tracking.

### Production reporting + provider compensation
Per-doctor production reports, technician productivity, compensation calculations integrated with the foundation payroll module.

## Frequently asked questions

### How does ERPClaw compare to AVImark or Cornerstone?
AVImark and Cornerstone are mature veterinary practice-management systems with deep imaging and lab equipment integrations. ERPClaw covers patient records, scheduling, treatment plans, vaccine reminders, inventory, and the full back-office ERP. For a clinic that wants one system end-to-end with AI-native architecture and $0 forever cost, ERPClaw is the AI-native option. For deep equipment-integrated workflows, AVImark and Cornerstone still lead on the operatory side.

### Does ERPClaw support multi-location clinics?
Yes. ERPClaw supports multi-warehouse inventory and per-location patient records. The shared database means a client whose pet visits Location A can have their records pulled at Location B without sync delay.

### Is the veterinary module free too?
Yes. healthclaw-vet is part of ERPClaw's open-source build. All modules including the 32 vet-specific actions are Free &amp; Open Source and $0 forever. Self-hosted on your hardware.

### What about controlled-substance tracking?
ERPClaw inventory supports controlled-substance audit trails through the immutable stock-ledger entry pattern. Every dispensing and administration creates an immutable SLE entry with chain-hashed history. State-specific reporting templates can be added per practice need.

Related: read the broader healthcare page, the dental version, the AI ERP buyer's guide, or the install guide.

---

## ERPClaw for small law firms: IOLTA-ready, AI-native, free
URL: https://www.erpclaw.ai/for/legal/small-firm/

# ERPClaw for small law firms
Matter records, IOLTA trust accounting, time tracking, billing, conflict checks, document management, payroll. AI-native, $0 forever.
41 legal-specific actions. 16 legal-specific tables.

## What ERPClaw does for small law firms
Small-firm legal practice management with matter records, client trust accounting (IOLTA), time tracking, billing (hourly + flat-fee + contingency), conflict checks, document management, and full back-office ERP.

## Capabilities

### Matter records + conflict checks
Per-matter records with practice area, opposing counsel, related parties, and tagged conflict-check fields. Pre-engagement conflict screen runs against the full client + party history before opening a new matter.

### IOLTA trust accounting
Per-client trust ledgers with three-way reconciliation (bank, trust ledger, client subledger). Bar-rule-aligned controls: no commingling, no overdraft, immutable trust GL with full audit trail. State-bar reporting templates ready.

### Time tracking + multi-rate billing
Time entries with billable/non-billable, matter, task code (LEDES if needed). Hourly billing, flat fee, contingency, hybrid arrangements. Pre-bill review workflow before invoicing the client.

### Document management with version control
Per-matter document folders, version history, and chain-hashed audit trail of every access. Self-hosted means privileged documents stay on your hardware.

### Calendar + deadline tracking
Court date entry, statute-of-limitations tracking, response-deadline calculations, and reminder workflows. The shared database means a deadline change in matter records reflects in the firm-wide calendar instantly.

## Frequently asked questions

### How does ERPClaw compare to Clio or PracticePanther?
Clio and PracticePanther are mature legal practice-management SaaS platforms with deeper court-rule calendars (CalendarRules integration) and broader e-filing connectors. ERPClaw covers matter records, IOLTA trust accounting, time tracking, billing, conflicts, document management, and the full back-office ERP. For small firms that want one system end-to-end with AI-native architecture and $0 forever cost, ERPClaw is the AI-native option.

### Is ERPClaw IOLTA-compliant for small-firm trust accounting?
ERPClaw enforces the technical controls bar rules require: per-client trust subledgers, three-way reconciliation (bank, trust ledger, client subledger), no commingling, no overdraft tolerance on trust accounts, and immutable GL with chain-hashed audit trail. Compliance is a function of your firm's policies plus the technical controls; ERPClaw provides the technical foundation. Self-hosted means client funds data stays on your hardware.

### Can ERPClaw handle contingency-fee accounting?
Yes. The Billing module supports contingency arrangements with per-matter fee structures. Recovery posts to client trust on receipt; allocation between firm fee, costs, and client distribution runs through the immutable GL. Year-end 1099 to clients (when applicable) integrates with the foundation finance module.

### Is the legal module free too?
Yes. legalclaw is part of ERPClaw's open-source build. All modules including the 41 legal-specific actions are Free &amp; Open Source and $0 forever. Self-hosted on your hardware.

Related: read the broader legal page, the AI ERP buyer's guide, the agency time-billing deep-dive, or the install guide.

---

## Small Business ERP - Enterprise Features at $0
URL: https://www.erpclaw.ai/for/small-business/

# ERP Without the
Enterprise Price Tag
The whole business in one system. 14 industries. Accounting, inventory, HR, manufacturing, CRM, and more. Enterprise-grade features that cost $0, set up in 5 minutes, and run from your chat app. Plus ERPClaw OS, your ERP improves itself, detects gaps, and suggests features.

## Small businesses deserve better
Traditional ERP was built for Fortune 500 budgets. You shouldn't need one to run your business properly.

### Traditional ERP Costs $50K+
SAP, Oracle, and NetSuite charge tens of thousands just for licensing. Add implementation consultants and you're looking at six figures before a single invoice is created.

### 6-18 Months to Go Live
Traditional ERP implementations drag on for months. Data migration, custom workflows, user training, go-live testing. Your business can't wait that long.

### Requires Dedicated IT Staff
Most ERP systems need a full-time admin to manage updates, troubleshoot issues, and configure new workflows. Small businesses don't have that luxury.

## How ERPClaw solves it
An AI-native ERP designed from scratch for businesses that want power without complexity.
Chat Interface, No TrainingJust type what you need in plain English. "Create an invoice for Smith & Co, $2,400 for web design." Your team is productive from day one.

5-Minute SetupInstall on your server, connect to Telegram or WhatsApp, and start running your business. No consultants, no migration projects, no months of configuration.

Self-Hosted, Free ForeverERPClaw is open source and runs on your own server. No per-user fees, no monthly subscriptions, no surprise price increases. Your data stays yours.

All-in-One: Every moduleAccounting, inventory, sales, purchasing, HR, payroll, manufacturing, CRM, projects, billing, and 14 industry verticals. One system replaces a dozen separate tools. Plus ERPClaw OS detects what your business needs and adapts.

## Built for businesses like yours
Real scenarios, real workflows. ERPClaw adapts to how you work.

### Retail Shop
A 12-person retail business tracks inventory across two locations, creates purchase orders when stock runs low, and generates sales invoices, all from a Telegram chat. Monthly close takes 10 minutes instead of two days.

### Service Company
A consulting firm with 25 employees manages project timesheets, expense claims, client invoicing, and payroll through one chat interface. The owner checks P&L and cash flow reports while commuting.

### E-Commerce Seller
An online seller uses ERPClaw to manage 500+ SKUs with batch tracking, automate purchase orders based on reorder levels, and reconcile payments. AI-powered anomaly detection flags unusual transactions before they become problems.

## Stop paying for software that holds you back
The whole ERP. 14 industry verticals. Unlimited users. Free forever. Self-host in 5 minutes.
Get Started Free See All Features

---

## Solo Founder ERP: Free Books for a Team of One
URL: https://www.erpclaw.ai/for/solo-founder/

# ERPClaw for Solo Founder
Built like a CFO; runs like a notebook.
All the ERP. None of the per-user fees. None of the SaaS bills.

## What you get

### Accounting + GL
Full double-entry, immutable, audit-ready

### Stripe Integration
every operation, ASC 606, MRR if you sell subscriptions

### Shopify Integration
every operation, payout reconciliation if you have a store

### CRM (light)
Customer profiles, deals, basic pipeline

### Selling / Invoicing
Send proper invoices instead of paypal.me links

### Reports
P&L, balance sheet, cash flow - all the standard ones

### AI-native chat
Type 'show me last month's P&L' and it appears

## Why Solo Founder teams switch

### $300/mo of SaaS subscriptions
QuickBooks + Stripe Sigma + Synder + DocuSign + a CRM = $300+/mo for a solo founder. ERPClaw is one tool, free.

### Setup is nights and weekends
You don't have time for a 45-day implementation. ERPClaw is install + sync + done in 5 minutes.

### Audit-ready before Series A
Investors will ask for clean books. ERPClaw's immutable GL is audit-ready from day one.

### Tools that grow with you
When you hire your first employee, payroll is in there. When you add inventory, it's in there. No tool migration.

## Real scenarios

### Pre-revenue founder
Bootstrap mode, no revenue yet. Track expenses, prep for first dollar. Free tool that scales.

### First-revenue Stripe founder
$5K MRR, paid landing page tool, no accountant. ERPClaw + Stripe = audit-ready from day one.

### Side hustle to full-time
Ecommerce side project growing into your main income. ERPClaw handles both phases without tool migration.

---

## Manufacturing Operations in ERPClaw
URL: https://www.erpclaw.ai/industries/manufacturing/

# Manufacturing operations, connected in one system.
Work orders, materials, and true costs. Free forever, self-hosted.
From bill of materials to finished goods with the books balanced the whole way: work orders, material consumption, outside processing, and finished items that carry their real cost instead of a guess.
Get ERPClaw on GitHubInstall guide

## Say what your business is. That is the setup.
Tell ERPClaw "I run a manufacturing business" and the manufacturing edition sets itself up: the records, workflows, and vocabulary of your business, on top of the same free ERP everyone gets. One install, no configuration project, and nothing forked away from the main product, so every improvement to ERPClaw is an improvement to your edition.
"Start a work order for 500 units of the compact model."

"Send the raw brackets to the coating shop and receive them back finished."

"What did a unit actually cost us to make last month?"

## What it runs for you

### Work orders that hold together
From plan to floor to finished goods, with material consumption recorded against the order it belongs to, not reconstructed at month end.

### Materials under control
Reservations that stop one order eating another's parts, putaway rules that route receipts to the right place, and ranked substitutes for the day the first-choice part runs short.

### Outside processing, honestly costed
Send materials to a subcontractor and receive finished goods whose cost includes the materials and the processing charge, posted once and balanced.

### Costs you can defend
Finished items carry real accumulated cost. When a customer pushes on price, you know your floor instead of estimating it.

### Quality and maintenance in the loop
Inspections and machine upkeep live beside production instead of in someone's notebook, so a failed check stops a shipment before the customer does.

### Planning from real demand
What to make and what to buy, derived from orders and stock on hand, in a system that already knows both.

## Want manufacturing updates as this edition deepens?
The software is free and installable today. Leave an email and you will hear when the manufacturing edition gains something worth your time. That is all the list is for.

All industry editionsERPClaw, the full ERPThe manufacturer's overview

---

## Retail Operations in ERPClaw
URL: https://www.erpclaw.ai/industries/retail/

# Retail operations, connected in one system.
Stores, stock, and margins. Free forever, self-hosted.
Run stores, stock, pricing, and the books behind them in one system, operated in plain language. Promotions that end when they should, loyalty that actually posts to the ledger, and a stock figure you can promise customers against.
Get ERPClaw on GitHubInstall guide

## Say what your business is. That is the setup.
Tell ERPClaw "I run a retail business" and the retail edition sets itself up: the records, workflows, and vocabulary of your business, on top of the same free ERP everyone gets. One install, no configuration project, and nothing forked away from the main product, so every improvement to ERPClaw is an improvement to your edition.
"Put the summer line on 20 percent off through Sunday, both stores."

"Which products made us money last month, and which just moved boxes?"

"How many of the blue chairs can I still promise, counting what is already sold?"

## What it runs for you

### Stores and channels
Multiple stores and online channels selling from the same stock truth, so a weekend sale in one place cannot quietly oversell another.

### Pricing and promotions
Price lists, promotions, and markdowns described in a sentence, applied consistently, and ended on time without a spreadsheet of exceptions.

### Stock you can promise
Live quantities with hard reservations underneath, so what the system says you can sell is what you can actually sell.

### Loyalty that reaches the books
Points and rewards are not a side app; they post where your accountant can see them, because the ledger and the counter are one system.

### Margins without exports
Ask what carried the quarter and get an answer from the books, item by item, with costs included, not a report you assemble by hand.

### The point of sale connection
Sales flow into the same books that run purchasing and stock, so end-of-day is a glance rather than a reconciliation.

## Want retail updates as this edition deepens?
The software is free and installable today. Leave an email and you will hear when the retail edition gains something worth your time. That is all the list is for.

All industry editionsERPClaw, the full ERPSelling online? The ecommerce view

---

## Schools Operations in ERPClaw
URL: https://www.erpclaw.ai/industries/schools/

# Schools operations, connected in one system.
Students, fees, and attendance. Free forever, self-hosted.
Enrollment, academics, attendance, staff, and fees in one system that also keeps real books, operated in plain language. Built for K-12 and higher education, with student records treated with the care regulations demand.
Get ERPClaw on GitHubInstall guide

## Say what your business is. That is the setup.
Tell ERPClaw "I run a school" and the schools edition sets itself up: the records, workflows, and vocabulary of your business, on top of the same free ERP everyone gets. One install, no configuration project, and nothing forked away from the main product, so every improvement to ERPClaw is an improvement to your edition.
"Enroll Priya Sharma into Grade 6, section B, starting Monday."

"Which families are behind on fees, and by how much?"

"Mark today's attendance for 6B: everyone present except two."

## What it runs for you

### Enrollment without the shoebox
Applications, admissions, and class placement as records that flow into everything else, instead of forms that get retyped three times.

### Fees that reach the ledger
Fee schedules, invoices to families, and payments recorded in real double-entry books, so the bursar and the accountant stop reconciling each other.

### Attendance in a sentence
Daily attendance recorded the way a teacher would say it, with the exceptions called out, and the term's pattern available when a parent asks.

### Academics and grading
Courses, grade entry, and transcripts that come from the system of record rather than a spreadsheet with one careful owner.

### Staff in the same system
Teaching and support staff records live beside payroll and the books, because a school is also an employer.

### Communication with a paper trail
Notices to families recorded against the student they concern, so the answer to 'were they told?' is a lookup, not a memory.

## Want schools updates as this edition deepens?
The software is free and installable today. Leave an email and you will hear when the schools edition gains something worth your time. That is all the list is for.

All industry editionsERPClaw, the full ERPThe education overview

---

## ERPClaw Insights: your AI analyst, no migration
URL: https://www.erpclaw.ai/insights/

# Your AI analyst. It knows your numbers so you don't have to dig for them.
Connect Stripe or Shopify and start asking: who owes us, where did the cash go, what is coming. Plain questions, straight answers, from real books. Free and open source, self-hosted, no migration, and nothing in your current setup gets touched.
Insights is set up per business, in order of request. Leaving your email here is the request.

## Ask it the way you'd ask a person
Real questions, real answers from the working engine, run on demonstration books. Nothing here is a mockup.
"How much do our customers still owe us, and how old is it?"
Copper Kettle Bakery$1,850.00 · 0 to 30 days
Harbor Lane Coffee$4,200.00 · 31 to 60 days
Oldest balance$975.00 · 91 to 120 days
Still owed to you$7,025.00

Demonstration books

"Are the books balanced right now?"
Money recorded out$2,000.00
Money recorded in$2,000.00
Booksbalanced

Demonstration books. Real accounting math, not a dashboard's opinion.

"Where did the cash go?"
Day-to-day operations-$2,000.00 · stocking up
Funding the business+$2,000.00
Net change$0.00

Demonstration books, with the line-by-line detail behind it

"The most expensive word in business software is migration. We built Insights so you never have to say it: connect what you already run, ask your first question the same day, and every system you rely on stays exactly where it is."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## What it does for you
The first group is on in every install. The second is switched on per business, so the setup matches your systems.
In every install

### Know who owes you
Ask and get the list: which customers, how much, and how long it has been sitting. No report builder, no export.

### Know what you owe
Your bills, by supplier and by age, so nothing surprises you on a Friday.

### Check the books balance
One question tells you whether the books are right, answered by real double-entry accounting underneath, not a dashboard's guess.

### See where the cash went
Money in, money out, and what it went to, from the actual ledger.

### Your business at a glance
Revenue, cash, what customers owe you, and whether you can cover what is coming, in one view.

### Catch strange activity
Unusual charges and odd movements get flagged, and it shows you exactly why each one was flagged so you can judge it yourself.

### See your cash weeks ahead
A best, expected, and worst view of your cash position, built from what customers owe you and what you owe suppliers.

### Chase late payers automatically
Overdue invoices get reminder emails automatically, on a schedule you set, in your words. The emails really go out.

### Connect Stripe in minutes
Your sales, fees, refunds and payouts flow into real books, history included. Live on the Stripe marketplace.

### Connect your Shopify store
Orders, refunds and payouts from your store land in the same books, cost of goods included.

### Bring bank transactions in
Hand it the statement file your bank already gives you and the lines match themselves to your invoices and payments.

### Everything sorted to the right place
Your Stripe and Shopify activity is mapped to the right accounts automatically, and you can see the rule that did it.

Switched on per businessAsk through the form below and we set yours up.

### Your week, in one email
A short Monday email in plain language: what changed, what needs you, what to watch.

### Live bank feeds
Your bank transactions arriving on their own, no statement file needed.

### Auto-sync with QuickBooks
Keep running QuickBooks. Insights reads your QuickBooks data and posts the matching journal entries back, with your approval, so both stay current.

### Auto-sync with NetSuite
Same shape for NetSuite: your system stays where it is, Insights reads its data and posts the matching journal entries back, with your approval.

### Auto-sync with Xero
And for Xero: your books stay put, Insights reads them and posts the matching journal entries back, with your approval.

### Automated month-end close
The software runs the monthly checking, matching, and tidying of the books, and brings you only the exceptions to approve.

## Three things the others don't do

### Real books underneath
Most AI money tools read a copy of your numbers and answer from the copy. Insights keeps a real set of double-entry books underneath, so the answer and the truth are the same thing.

### It does things, not just charts
A dashboard shows you the overdue invoice. Insights sends the reminder. Answers are good; work taken off your plate is better.

### It grows with you
The day you want more than answers, the full ERPClaw business system is already under your data. Flip the switch; there is nothing to migrate. And it is free forever, self-hosted, so growing costs nothing.

Comparing tools side by side? See the head-to-heads:Campfire ·Rillet ·DualEntry ·Puzzle ·Billow

## Start with answers. Keep the door open.
Insights reads and advises. The day you want it to keep the books, run the invoicing, the stock, the payroll, that is ERPClaw: same install, flip the switch, nothing to migrate because your data is already there. If your business has its own vocabulary, there is probably an edition that already speaks it.

## Common questions

### What is ERPClaw Insights?
Your AI analyst. Connect Stripe or Shopify and get plain-language answers about cash, customers and costs. No migration, nothing to replace. Out of the box it reads and advises; it changes nothing in the systems you run unless you turn on the opt-in sync.

### Why are some features switched on per business?
Because they hook into systems that differ from business to business: banks, QuickBooks, NetSuite, Xero. We switch each one on per business so the setup matches your systems. Ask through the form on this page and we reach out to set yours up.

### Will it mess with my QuickBooks or NetSuite?
Not unless you ask it to. Out of the box, Insights only reads: your current books stay exactly as they are while it answers questions on the side. The auto-sync is the one exception, and it is opt-in: turn it on and Insights posts entries back into your system, with your approval, so both stay current.

### What does it cost?
ERPClaw is free forever, self-hosted, open source. Paid managed hosting, where we run it for you, is available by request through the pricing page.

### How is this different from the AI accounting tools I keep seeing?
Most of them either keep a copy of your numbers and report on it, or they are a service that does your books for a monthly fee on top of the software you already pay for. Insights is software you own, with real books underneath, and it feeds the same free system that can eventually run your whole business. See the head-to-head pages for Campfire, Rillet, DualEntry, Puzzle, and Billow if you are comparing.

### What happens when I want more than answers?
That is the quiet advantage. Your data already sits in a real set of books, not a reporting copy, so upgrading to the full ERPClaw is a switch to flip, not a migration project. Tools built as a layer on someone else's books cannot offer that.

### Can you just do my bookkeeping or taxes for me?
We make software, and the one service we offer ourselves is implementation help getting set up. For bookkeeping, tax, or close work done by people, our accounting partners can help; reach out through the support page and we will connect you.

### How do I get in?
Leave your email on this page and we reach out to set you up, in order of request.

## We set this up for you, one business at a time.
One form covers everything on this page: Insights itself, the per-business features above, and managed hosting if you want us to run it. Your email, nothing else, and we only write when there is something real to say.

---

## Bank feeds and statement import for ERPClaw
URL: https://www.erpclaw.ai/integrations/bank-feeds/

# Bank feeds for ERPClaw: your transactions, in your books
The shipped path is statement file import: download a file from your bank, hand it over in a sentence, and the lines match themselves to open invoices and payments. It works with the four formats banks actually export, and it is free, like the rest of ERPClaw.
Read the import guideAbout ERPClaw

## The four formats your bank already exports
No connector account, no monthly fee, no third party between your bank and your books. The format is detected for you.

### OFX
The common download from US and Canadian banks

### CAMT.053
The ISO 20022 statement most European banks issue

### MT940
The SWIFT format used across Europe, the Middle East, and Asia

### BAI2
The cash-management format US commercial banks provide

## Matching does the boring half
A deposit that clears an outstanding invoice pairs with that invoice. A withdrawal that settles a bill pairs with that payment. The lines the engine is confident about are matched in one pass, and standing match rules let your recurring lines, payouts, fees, subscriptions, categorise themselves. What is left is the short list a human should actually look at.
Ask "how does the checking account reconcile as of the end of January?" and you get the picture at a cutoff, from the books, in plain language. That is the AI-native difference: the request is the interface, and deterministic accounting code does the posting.

## Multi-currency, the honest version
Payments are where multi-currency is real and validated in ERPClaw: an invoice raised in a supported currency is settled in that same currency, and the books stay clean because no silent conversion ever happens inside a transaction. Cross-currency conversion at posting time is deliberately not what ERPClaw does; that discipline is what keeps a self-run ledger defensible.

## Common questions

### How do I get bank transactions into ERPClaw today?
Download a statement file from your bank in OFX, CAMT.053, MT940, or BAI2 and hand it to ERPClaw. The lines import, match themselves to open invoices and payments where the engine is confident, and queue the rest for your review. This is shipped and documented.

### Does re-importing the same statement create duplicates?
No. Every line carries its own identifier from the bank, and lines already on the books are skipped on the way in, so an interrupted download can be run again safely.

### Can ERPClaw learn my recurring transactions?
Yes. Match rules let recurring lines categorise themselves, for example anything with STRIPE in the description mapping to your Stripe payout account, with priorities deciding which rule wins.

### Is there a live bank connection instead of files?
Yes. Live feeds are connected per business so the setup matches your bank; ask through the form at the bottom of this page and we set yours up. Statement file import needs no setup at all and works today.

### What about payments in other currencies?
Payments in the currencies ERPClaw supports settle and reconcile cleanly, with each invoice paid in its own currency. ERPClaw deliberately does not convert currencies inside a transaction; an invoice is settled in the currency it was raised in.

### What does this cost?
Nothing. ERPClaw is free forever, self-hosted, and the statement import is part of it.

## We set this up for you, one business at a time.
Live bank feeds are connected per business so the setup matches your bank. Leave your email and we set yours up. Statement file import needs no setup at all: it works today, free.

---

## NetSuite import: move from NetSuite to ERPClaw
URL: https://www.erpclaw.ai/integrations/netsuite-import/

# NetSuite import: off the meter, onto your own machine
NetSuite by Oracle is capable software with a renewal problem. ERPClaw is the AI-native ERP with no renewal at all: free forever, self-hosted, PostgreSQL or SQLite, operated in plain language. The move runs on exports NetSuite already gives you.
Compare ERPClaw and NetSuiteInstall ERPClaw first

## The migration path that works today
File-based, shipped, documented. No vendor call required to leave a vendor.

### Master records, by CSV
Customers, vendors, items, and your sales pipeline export from NetSuite as CSV and import into ERPClaw with explicit duplicate handling, in transactions that fully land or fully roll back.

### Bank history, from the bank
Skip NetSuite as the middleman: statement files in OFX, CAMT.053, MT940, or BAI2 come straight from your bank and match themselves to open invoices and payments.

### Multi-entity from day one
Consolidation groups, intercompany transactions and re-runnable eliminations are shipped in ERPClaw. Translating a foreign subsidiary is a deliberate entry your accountant makes rather than a number the system invents. Payments settle in the currency the invoice was raised in, and no transaction ever hides a silent conversion.

Bank statement import guideCSV import and export guideReports by department or project

## Why AI-native matters at this size
Mid-market NetSuite shops often carry dedicated admins because the interface is the bottleneck: saved searches, custom forms, permissions. ERPClaw was built AI-native from the start, so "show me the P&L for June grouped by department" is a request, not a report-building exercise. The AI decides what you meant; deterministic accounting code decides what the books say; an invariant engine checks every posting. Bolting a chat window onto a legacy ERP cannot reproduce that, because the architecture underneath is still menus.

## Two ways in
The file-based path works today and needs nobody's permission: exports out of NetSuite, imports into ERPClaw, opening balances, done. The direct importer is the second way: it reads your NetSuite account and rebuilds transaction-level history end to end. NetSuite accounts differ enormously, saved searches, custom records, years of history, so we enable the direct importer per business, matched to your account; the form below is how you ask.

## Common questions

### Why would anyone leave NetSuite by Oracle for a free ERP?
The bill and the admin load. NetSuite pricing is negotiated, renewals climb, and specialist administration is its own cost. ERPClaw is $0 forever, self-hosted, and operated in plain language rather than through an admin console.

### Can ERPClaw really cover what NetSuite covers?
For the functions most businesses actually use, the books, inventory, purchasing, manufacturing, sales, billing, and multi-entity consolidation, yes, in one install. What ERPClaw does not do is charge per module or per seat.

### Is there a direct NetSuite importer?
Yes, enabled per business so the migration matches your account and customizations; request it at the bottom of this page and we set you up. The file-based path needs nothing and works today: CSV for master records, statement files for bank history, opening balances to start clean.

### How does multi-currency work?
Payments are the validated multi-currency path: invoices settle in the currency they were raised in, across the currencies ERPClaw supports. Consolidation eliminates intercompany activity, and translating a subsidiary is recorded deliberately rather than computed for you. ERPClaw deliberately never converts currency inside a transaction.

### What happens to our customizations?
They do not port, and honestly, most exist to bend NetSuite toward how you already talk about your business. ERPClaw's interface is plain language over your own vocabulary, which removes the reason many of those customizations existed.

### What does migration cost?
The software costs nothing, forever. Your costs are your own time and hosting. There is no implementation fee because there is no vendor invoice at all.

## We set this up for you, one business at a time.
The direct NetSuite importer is enabled per business so the migration matches your account. Leave your email and we set yours up. The file-based path needs nothing: it works today, free.

---

## QuickBooks import: move from QuickBooks to ERPClaw
URL: https://www.erpclaw.ai/integrations/quickbooks-import/

# QuickBooks import: bring your books to a system you own
QuickBooks Online by Intuit is where many businesses start. ERPClaw is where the subscription ends: the AI-native ERP, free forever, self-hosted, with your data in a database you can open yourself. The move runs on files you can export from QuickBooks today.
Compare ERPClaw and QuickBooksInstall ERPClaw first

## The migration path that works today
Three file-based routes, all shipped, all documented, all free.

### Your lists, by CSV
QuickBooks exports customers, vendors, and items as spreadsheets, and ERPClaw imports CSV with explicit duplicate handling, so your master records come across without retyping.

### Your bank history, by statement file
Bank transactions come straight from the bank in OFX, CAMT.053, MT940, or BAI2, and match themselves to open invoices and payments. No connector fee, no third party.

### Your sales pipeline
Leads, contacts, companies, and deals import from CSV into ERPClaw's built-in Sales, each import running as one transaction that either fully lands or fully rolls back.

Bank statement import guideCSV import and export guide

## What you gain on the other side
QuickBooks with AI features added is still QuickBooks: menus first, AI decorating the edges. ERPClaw is AI-native, built so the plain-language request is the interface and deterministic accounting code does the posting. And it is a whole ERP, so the inventory, purchasing, and sales pipeline that QuickBooks hands off to other subscriptions live in the same books here.

## Two ways in
The file-based path above works today and businesses complete it on their own: exports out of QuickBooks, imports into ERPClaw, opening balances, done. The one-step importer is the second way: it reads your QuickBooks account directly and rebuilds your records without the export step. Because every QuickBooks setup is a little different, we enable it per business, matched to yours; the form below is how you ask.

## Common questions

### Can I move from QuickBooks Online by Intuit to ERPClaw today?
Yes, two ways. The file-based path works today: export your lists from QuickBooks as CSV and your bank history as statement files, import both into ERPClaw, and set your opening balances. The one-step importer that reads your QuickBooks account directly is enabled per business; request it at the bottom of this page and we set you up.

### What does ERPClaw cost after QuickBooks?
Nothing, forever. ERPClaw is open source and self-hosted. There is no subscription to replace the one you are leaving.

### Do I lose my history?
Your QuickBooks account and its exports remain yours, and what you import into ERPClaw is yours in a database on your own machine. ERPClaw never holds your data hostage; export is a first-class action.

### Is ERPClaw actually more capable than QuickBooks?
They are different shapes. QuickBooks is accounting software. ERPClaw is a full AI-native ERP: the books plus inventory, manufacturing, purchasing, and built-in Sales, operated in plain language, in one system.

### What is the catch with a free ERP?
You host it yourself, on your own machine or server. That is the trade: your infrastructure, your data, no bill.

### How do I ask questions about the imported data?
In plain language. 'Which customers owe us the most?' or 'what did we make last quarter?' are requests, not report-builder projects, because the AI-native interface is the product.

## We set this up for you, one business at a time.
The one-step QuickBooks importer is enabled per business so the migration matches your setup. Leave your email and we set yours up. The file-based path needs nothing: it works today, free.

---

## Xero Sync and Migration: Connect or Move for Free
URL: https://www.erpclaw.ai/integrations/xero/

# Xero and ERPClaw: keep it in sync, or leave it behind
Two honest paths for Xero businesses. Keep Xero and let ERPClaw sync with it, reading your data and posting the matching journal entries back with your approval. Or move to ERPClaw entirely and stop paying for accounting software at all, because ERPClaw is free forever, self-hosted, and runs in plain language.
Keep Xero, add the AI analystInstall ERPClaw

## The move, in three file-based steps
Works today, documented, free, and yours to run. No vendor call required to leave a vendor.

### Your lists, by CSV
Xero exports contacts and items as spreadsheets, and ERPClaw imports CSV with explicit duplicate handling, so your master records come across without retyping.

### Your bank history, from the bank
Statement files in OFX, CAMT.053, MT940, or BAI2 come straight from your bank and match themselves to open invoices and payments. No connector fee, no third party.

### Opening balances, then go
Set your opening balances and the books start clean, with history where you need it and a clear line where ERPClaw takes over.

Bank statement import guideCSV import and export guideWhat your accountant will ask about

## Or keep Xero, and stop retyping
The Xero sync is connected per business, matched to your organisation's setup: ERPClaw reads your Xero data and posts the matching journal entries back, with your approval, so both stay current without anyone copying rows between tabs. Pair it with ERPClaw Insights and your Xero books gain a plain-language analyst on top: who owes you, where cash went, what is coming.

## Common questions

### Can ERPClaw sync with Xero automatically?
Yes. The Xero sync is connected per business so it matches your organisation's setup: ERPClaw reads your Xero data and posts the matching journal entries back, with your approval, keeping both current. Ask through the form at the bottom of this page and we set yours up.

### Can I move off Xero completely?
Yes, by files, today: export your contacts and items as CSV, bring bank history in as statement files, set opening balances. The whole path is documented and free, like the rest of ERPClaw.

### Does ERPClaw cover my region like Xero does?
UK and EU regional coverage is built in, alongside US, Canada, and India. If you run in Australia or New Zealand, Xero's local depth there is real; compare carefully for those two.

### What about multi-currency?
Payments settle in the currency each invoice was raised in, across the currencies ERPClaw supports. ERPClaw deliberately never converts currency inside a transaction, which keeps a self-run ledger defensible.

### What does it cost after Xero?
Nothing, forever. ERPClaw is open source and self-hosted, with no subscription to replace the one you are leaving. If you would rather not host it, Cloud Managed is our run-it-for-you option, by request through the pricing page.

### Is my accountant going to hate this?
Show them the books: real double-entry, immutable postings with visible reversals, and a full audit trail. If you also want human bookkeeping or tax help, our accounting partners can help; ask through the support page.

## We set this up for you, one business at a time.
The Xero sync is connected per business so it matches your organisation. Leave your email and we set yours up.

---

## Migrate to ERPClaw: from QuickBooks, NetSuite, Xero, Odoo
URL: https://www.erpclaw.ai/migrate/

# Migrate to ERPClaw
Pick your source system. The AI agent handles the transcription. Most teams cut over in a weekend with a one-week parallel run. $0 forever after.

## Live migration guides

### From QuickBooks
QBO Plus, QBO Advanced, or QuickBooks Desktop. Most common path.
Common pain: Hit the user cap. Audit-trail concern. Manual Stripe reconciliation.

### From NetSuite
Enterprise SuiteCloud. Renewal shock or implementation rage.
Common pain: 30-40% renewal increase. SuiteScript lock-in. Two FTEs just on admin.

### From Odoo
Community or Enterprise exports into ERPClaw modules.
Common pain: Edition boundaries, custom modules, and recurring per-user cost.

### From ERPNext
Move from Frappe documents into ERPClaw actions and records.
Common pain: Custom app dependencies and framework-specific workflows.

### From SAP Business One
Plan the move from HANA or SQL Server with a controlled cutover.
Common pain: Partner dependency, custom add-ons, and administration overhead.

## The migration approach
Every migration follows the same shape. Export from the source system. AI agent reads the export and runs ERPClaw's import actions to recreate the data. You spot-check trial balance and aging reports. Run parallel for a week. Cut over.
The AI piece matters. ERPClaw is AI-native: every action that creates a customer, posts a journal entry, or imports a transaction is callable from a chat agent. You don't write a migration script. You ask an agent to do the import, and it runs the right actions for your data.
The data integrity piece matters. ERPClaw enforces accounting rules checked as each entry is written: double-entry balance, voucher balance, immutability of submitted GL entries, invariant checks. If the source system had quiet data errors, the migration will surface it. That's a feature.

### What the AI cannot do for you

- ·Pick your chart of accounts. ERPClaw seeds a standard US GAAP chart; you decide whether to keep it or adapt yours.
- ·Decide your fiscal year close date. ERPClaw asks; you answer.
- ·Re-key custom integrations the AI can't reverse-engineer (rare, but happens with deeply customized SuiteScript on NetSuite).

## Moving off something else?
Sage Intacct, Xero, Odoo, ERPNext, FreshBooks, or anything else: the same file-based path applies, and we will walk yours with you. Reach us through support and we size the move against your actual system.

### From Sage Intacct
Mid-market financial SaaS

### From Xero
SMB cloud accounting

### From Odoo
Open-core ERP, Community or Enterprise

### From ERPNext
Open-source ERP

### From SAP Business One
SMB ERP on HANA or SQL Server

### From FreshBooks
Agency time-billing

Migration directory

## Choose your migration path
Each guide focuses on the source system, the information to bring across, and the checks to run before cutover.
From QuickBooksFrom NetSuiteFrom OdooFrom ERPNextFrom SAP Business One

## Frequently asked questions

### How long does an ERPClaw migration actually take?
From clean QBO export to ERPClaw running with the same data takes most teams a few hours of work spread across a weekend. The AI agent handles the transcription so you don't manually re-key journal entries. NetSuite migrations take longer because SuiteScript customizations have to be re-expressed as ERPClaw actions, but the AI assists that translation.

### Will I lose my historical data?
No. The migration preserves your full transaction history. The agent imports chart of accounts, customers, vendors, items, and historical journal entries from your source system. You can run year-over-year comparison reports the day after migration. Your archived books in the source system stay readable for as long as that vendor lets you.

### Do I need a developer or consultant to migrate?
Many teams have an AI assistant do the heavy lifting of the export-and-import work; ERPClaw ships the specifications those assistants read, so the move is well-marked. ERPClaw has SKILL.md specs per module; the AI reads those plus your source export and runs the import actions. AvanSaber Inc. offers paid implementation services for teams that prefer not to self-host.

### What happens during the migration weekend?
Friday: install ERPClaw on your hardware, export from your source system. Saturday: AI agent transcribes the data, you spot-check the trial balance and AR/AP aging against the source. Sunday: parallel run for one week to catch any reconciliation gaps. Monday following: cut over fully. No firm needs a 6-month consulting engagement.

### What's the cost of migrating off SaaS to self-hosted ERPClaw?
Software cost: $0 forever. Server cost: your choice ($0 if you self-host on existing hardware, $5 to $20 per month for a VPS if you prefer cloud). Implementation: $0 if you self-migrate with an AI agent, or paid services from AvanSaber Inc. if you want managed help. Compare that to the all-in cost of staying on your current SaaS stack.

Related: read the AI-native ERP framework, the 13-vendor matrix, the pricing page ($0 forever), or per-source comparisons at /compare/quickbooks/ and /compare/netsuite/.

---

## Migrate from ERPNext to ERPClaw
URL: https://www.erpclaw.ai/migrate/from-erpnext/

# Migrate from ERPNext to ERPClaw
Open source to open source. AI-assisted DocType transcription, schema mapping, cutover. Single-stack install. Both Free & Open Source.

## Why teams move from ERPNext
ERPNext by Frappe Technologies is a respected open-source ERP with 15 years of history, but the operational and AI stories both create friction. The Frappe stack requires MariaDB or PostgreSQL plus Redis plus Node.js plus Nginx plus Bench, each with its own install and upgrade lifecycle. AI capabilities depend on third-party Frappe apps like NextAI or ChatNext, which fragment install and version management. Install and upgrade pain is a common community complaint.
ERPClaw is AI-native at the action layer, single-stack (Python plus SQLite or PostgreSQL), Free & Open Source, self-hosted, and $0 forever.

- ·Single-stack install. Python plus one database. No Redis, Node, Nginx, or Bench.
- ·AI-native, not bolt-on. Every action callable from a chat agent, no third-party AI app required.
- ·12-step GL validation. Constitutional invariants enforced at write time.
- ·Database-agnostic. SQLite default or PostgreSQL via PyPika.

## What ERPClaw replaces
DocType-to-module mapping. Most ERPNext modules have a direct ERPClaw equivalent.
ERPNext
Accounts (Chart, Journal Entry, GL)

ERPClaw
erpclaw foundation: the core tables, full GL invariant validation, immutable posted entries

ERPNext
Selling (Quotation, Sales Order, Sales Invoice)

ERPClaw
erpclaw foundation sales plus erpclaw-integrations-stripe

ERPNext
Buying (RFQ, Purchase Order, Purchase Invoice)

ERPClaw
erpclaw foundation purchasing module

ERPNext
Stock (Warehouses, Stock Entry, Stock Ledger)

ERPClaw
erpclaw foundation stock module with immutable lineage

ERPNext
CRM (Leads, Opportunities)

ERPClaw
erpclaw-growth CRM

ERPNext
Manufacturing (BOM, Work Order, MRP)

ERPClaw
erpclaw-ops manufacturing module

ERPNext
HR (Employee, Leave, Attendance)

ERPClaw
erpclaw foundation HR module

ERPNext
Projects (Project, Task, Timesheet)

ERPClaw
erpclaw-ops projects

ERPNext
Assets (Asset, Depreciation)

ERPClaw
erpclaw-ops assets

ERPNext
Support (Issue, Service Level Agreement)

ERPClaw
erpclaw-ops support

ERPNext
Quality Management

ERPClaw
erpclaw-ops quality

ERPNext
Healthcare

ERPClaw
healthclaw plus dental, vet, mental, homehealth subverticals

ERPNext
Education

ERPClaw
educlaw plus k12, lms, finaid, statereport, scheduling, highered subverticals

## What ERPNext does better
Honest disclosure. Both projects are open source and we respect the work. If any of these are core to your business, weigh them before committing.

### Frappe DocType system is mature
Frappe's DocType abstraction has 15 years of polish. Define a DocType in the UI and you get list views, form views, permissions, and REST API endpoints for free. ERPClaw's spec-first approach is different: you describe what you want in a SKILL.md spec and an AI agent generates the action. Different model, different audience.

### Multi-tenant per-site model
ERPNext's Frappe framework supports multiple sites per server with isolated databases. ERPClaw runs one ERP per install. If you host ERPNext for multiple separate customer companies on one server, that pattern maps to one ERPClaw install per company, which the $0 license makes a cost-free difference.

### Large existing community
ERPNext has 15 years of community contributions, regional partners worldwide, and a deep base of third-party apps on Frappe Cloud. ERPClaw is a newer project with a smaller but growing community. If you rely on a specific Frappe app or regional partner, verify the ERPClaw equivalent before committing.

### Polished web UI for non-technical users
ERPNext ships a full web UI for daily operations. ERPClaw is action-first: the canonical interface is the AI agent and CLI; webclaw is dev-only for now. If your finance and operations team need a polished click-through UI today, that is the real constraint.

## How the migration works

- 1

### Export from ERPNext

ERPNext exports cleanly per DocType. Use Bench (the bench export-doc command with a site flag), or use the Report Builder UI to export each DocType to CSV. Pull Account, Customer, Supplier, Item, Journal Entry, Sales Invoice, Purchase Invoice, Stock Ledger Entry, and your custom DocTypes. Save each as CSV in erpnext-export-YYYYMMDD/. Take a MariaDB or PostgreSQL dump as your safety net.

- 2

### Map DocTypes to ERPClaw modules

Walk an AI agent through the ERPNext to ERPClaw field map. Account becomes ERPClaw account, Customer becomes customer, Item becomes item, Journal Entry rows become gl_entry rows. The agent reads SKILL.md specs for each ERPClaw module and produces a mapping plan. Custom DocTypes need explicit decisions: which ones map to existing ERPClaw modules, which need a new module generated from spec.

- 3

### Install ERPClaw and run the data transfer

Follow /docs/core/install/. ERPClaw is single-stack: Python plus SQLite (default) or PostgreSQL via PyPika. No MariaDB, no Redis, no Node.js, no Nginx, no Bench. Point the AI agent at the export folder and your fresh ERPClaw install. The agent calls setup-chart-of-accounts, add-account, add-customer, add-supplier, add-item, and post-gl-entries to recreate your books. Frappe Server Scripts and Client Scripts need a separate AI conversation per script.

- 4

### Verify and cut over

Run check-gl-integrity in ERPClaw. Compare trial balance, AR aging, AP aging, stock valuation against the ERPNext reports you exported. Investigate any drift before going live. Run parallel for one to two weeks, then cut over. ERPNext is free so timing is less critical than for a SaaS subscription; you just stop using it on cutover day.

Realistic timeline: 3 to 6 weeks elapsed for a team with vanilla ERPNext. 6 to 10 weeks for teams with heavy DocType customization and many Server Scripts.

## Stack comparison

### ERPNext + Frappe

- ·Python 3.10+
- ·Frappe framework
- ·MariaDB or PostgreSQL
- ·Redis (cache, queue, socketio)
- ·Node.js (for build tooling and socketio)
- ·Nginx (production)
- ·Bench (process manager and CLI)

### ERPClaw

- ·Python 3.10+
- ·SQLite (default) or PostgreSQL
- ·No Redis required
- ·No Node.js required
- ·No Nginx required for core
- ·No separate process manager

The single-stack install matters most for self-hosted teams running their own infrastructure. Fewer components means fewer upgrade paths, fewer security patches to track, and faster recovery when something breaks.

## Frequently asked questions

### Can I export my ERPNext data?
Yes, cleanly. Every DocType in ERPNext exports to CSV via Bench (bench export-doc) or the Report Builder UI. You can also take a full MariaDB or PostgreSQL dump as a safety net. The AI agent reads the CSV exports and runs ERPClaw import actions; you do not write a migration script by hand.

### Will my custom DocTypes port?
Mapping is straightforward for custom DocTypes that mirror the common ERP modules (custom Customer or Item extensions, for example). Vertical-specific custom DocTypes need module-level work: the AI agent can generate a new ERPClaw module from spec if your custom DocType is not covered by an existing module. Budget extra time if you have 10 or more custom DocTypes that are not simple field extensions.

### Both ERPClaw and ERPNext are Free &amp; Open Source and open source. Why switch?
Three structural reasons. First, ERPClaw is AI-native at the action layer: every action is callable from a chat agent, with no third-party AI app. ERPNext relies on third-party Frappe apps like NextAI or ChatNext for AI, which adds install and version complexity. Second, ERPClaw is single-stack: Python plus SQLite or PostgreSQL. No MariaDB, Redis, Node.js, Nginx, or Bench. Third, ERPClaw enforces 12-step GL validation at write time and invariant checks across the system. ERPNext has accounting validation but the constitutional invariant pattern is unique to ERPClaw.

### What about my Frappe Server Scripts and Client Scripts?
Server Scripts and Client Scripts need re-implementation in ERPClaw. The good news: ERPClaw's spec-first regen pattern means an AI agent can generate the new action from a SKILL.md description. Tell the agent what each script does and it produces the ERPClaw equivalent. Most server-side logic ports cleanly; complex Frappe-specific framework calls need more care.

### How long does the migration take?
For a team with vanilla ERPNext and few custom DocTypes, expect 3 to 6 weeks elapsed including the parallel run. For teams with heavy DocType customization and many Server Scripts, expect 6 to 10 weeks. The AI agent handles bulk data transcription quickly; the customization translation is the variable cost.

### Is ERPClaw really free given ERPNext is also free?
Yes, both are free at the software level. The forward cost difference is in operational complexity: ERPNext requires you to operate MariaDB or PostgreSQL plus Redis plus Node.js plus Nginx plus Bench. ERPClaw requires only Python plus your chosen database. For self-hosted teams, that is a meaningful operational cost reduction. For Frappe Cloud users, ERPClaw self-hosted is genuinely $0 versus a Frappe Cloud subscription.

### What about Frappe Cloud or hosted ERPNext?
Frappe Cloud and other hosted ERPNext providers charge monthly subscriptions per site. ERPClaw self-hosted is $0 forever. If you currently pay for Frappe Cloud or a managed ERPNext provider, the migration is also a cost reduction. AvanSaber Inc. offers paid implementation services if you prefer managed help; the cost is far below ongoing managed-ERPNext fees.

---

## Migrate from NetSuite to ERPClaw: mid-market playbook
URL: https://www.erpclaw.ai/migrate/from-netsuite/

# Migrate from NetSuite to ERPClaw
Mid-market playbook. AI-assisted data transcription, SuiteScript translation, two-week parallel run. Cancel at renewal. $0 forever after.

## Why teams move
NetSuite renewal increases of 30 to 40 percent are widely reported. Implementation cost was already $50K to $500K. Two full-time admins are commonly required just to operate the tenant. Customer reports of "$30K projected, much more in reality" are consistent. ERPClaw is $0 forever, AI-native, open-source, with the same primitives at the mid-market entry point.

- ·No license fee. ERPClaw is Free & Open Source.
- ·No SuiteScript lock-in. Open source, spec-first regen, action layer is the API.
- ·No 6-month implementation. Most mid-market migrations are 2 to 6 weeks elapsed.
- ·No 1-2 admin FTEs. Self-host on your existing infrastructure.
Not ready to move the books yet? AI for NetSuite covers the middle path: keep NetSuite and put a plain-language AI analyst beside it.

## 8-step playbook

- 1

### Inventory your SuiteScript and SuiteCloud customizations

Before exporting data, list every script, workflow, saved search, and SuiteCloud bundle that touches your books. The AI agent helps re-express each as an ERPClaw action. Heavy customization (50+ scripts) needs more migration time; vanilla NetSuite (no SuiteScript) migrates faster.

- 2

### Export from NetSuite

Reports → Trial Balance, Customer/Vendor list, Item list, Chart of Accounts, full GL detail (per period), AR aging, AP aging, Saved Searches as CSV. NetSuite exports cleanly to CSV; the AI agent handles the import.

- 3

### Install ERPClaw on your hardware or VPS

Follow /docs/core/install/. PostgreSQL is recommended for NetSuite migrations because the historical data volume is usually significant. Same code targets either backend via PyPika.

- 4

### Talk to the AI agent

Point an AI agent at the NetSuite export and your fresh ERPClaw install. The agent reads SKILL.md specs, maps NetSuite fields to ERPClaw actions (setup-chart-of-accounts, add-account, post-gl-entries, add-customer, add-vendor, add-item, etc.), and runs the import. SuiteScript translation is a separate AI conversation; tell the agent what each script does and it generates the ERPClaw equivalent.

- 5

### Spot-check trial balance and aging

Run check-gl-integrity. Compare ERPClaw trial balance against the NetSuite trial balance you exported. They should match. AR aging buckets should match. AP aging buckets should match. Investigate any drift before going further.

- 6

### Translate SuiteScript customizations

For each SuiteScript or workflow that touches your books, decide: keep, replace with an ERPClaw action, or remove. ERPClaw's spec-first regen pattern means the AI generates the new action from a SKILL.md description. This is where most NetSuite migration time goes.

- 7

### Run parallel for two weeks

NetSuite migrations get a longer parallel run because the customization surface is larger. Post in both systems. Reconcile every day. By week two, the answer should be obvious: cut over or kick the migration plan back.

- 8

### Cancel NetSuite at renewal

ERPClaw is free forever. Time the cutover so you don't pay another full year on NetSuite. Most teams plan their migration to land 60 days before NetSuite renewal so the parallel run + buffer fits cleanly.

## Common NetSuite pain, fixed in ERPClaw

### 30 to 40 percent NetSuite renewal increase
ERPClaw is $0 forever. The renewal does not happen; you cancel.

### $50K to $500K NetSuite implementation cost
ERPClaw self-migration is free if you run it yourself. AvanSaber Inc. offers paid implementation services if you prefer managed help; the cost is far below NetSuite implementation.

### Two full-time admins needed for NetSuite
ERPClaw is self-hosted with a 5-minute install and SQLite or PostgreSQL backend you may already run. No NetSuite-specialist FTE required.

### SuiteScript lock-in
ERPClaw is Free &amp; Open Source with SKILL.md specs. Custom logic is regenerable from spec by any LLM. No proprietary scripting language; the action layer is the API.

### Vendor cloud lock-in (Oracle SaaS)
ERPClaw is self-hosted. Your data lives on your hardware in a SQLite file or PostgreSQL database. The export pattern is literally cp data.sqlite somewhere-else.sqlite.

### Multi-month implementation timeline
ERPClaw migrations are typically 2 to 6 weeks total elapsed time including parallel run. The actual hands-on work is days to a couple of weeks for most mid-market teams.

## Frequently asked questions

### How long does a NetSuite-to-ERPClaw migration actually take?
It depends on your SuiteScript surface. A vanilla NetSuite tenant (no scripts, no SuiteCloud bundles) migrates in 2 to 3 weeks elapsed including the parallel run. A heavily customized tenant with 50+ scripts can take 6 to 8 weeks. The AI agent handles the data transcription quickly; the SuiteScript translation is the variable cost.

### Can ERPClaw handle our multi-entity setup?
Yes for most mid-market multi-entity setups. ERPClaw supports multiple companies in a single shared database with per-entity GL, multi-currency at 7 currencies, and entity-level reporting. Where ERPClaw is honestly different from NetSuite OneWorld is transactional FX revaluation at scale, which is intentionally out of ERPClaw's scope, not a roadmap item, and automatic consolidation-time translation, which is recorded manually today.

### What about my SuiteApps?
Most SuiteApps are vendor wrappers around standard ERP primitives. The AI agent helps re-express each SuiteApp's logic as ERPClaw actions or vertical modules. ERPClaw ships 14 native verticals (healthcare, education, construction, etc.) plus the foundation; the gap surface is usually narrower than NetSuite admins expect.

### Will my CFO accept moving off NetSuite to open-source ERPClaw?
Yes if you frame it correctly. The open-source argument is the same argument for Linux on production servers: observable controls beat SaaS black-box controls for audit defense. ERPClaw has constitutional rules enforced on every operation, layered automated testing, an immutable ledger, a chain-hashed audit trail, and a 12-step posting validation. The NetSuite admin TCO is a separate material reduction. CFOs are increasingly comfortable with this stack.

### What about NetSuite's vendor-managed support?
ERPClaw support is community-tier (GitHub issues, Discord) plus optional paid services from AvanSaber Inc. The math: 1 NetSuite admin FTE costs roughly $120K per year all-in, which buys a lot of paid AvanSaber support. The trade-off is real but not punitive at the mid-market.

### Can I run ERPClaw at NetSuite-scale data volumes?
PostgreSQL backend handles tens of millions of GL entries cleanly. The action layer is designed for AI agents to drive bulk operations. Where ERPClaw is honestly weaker today is single-page web dashboards for non-technical users; webclaw is dev-only. If your finance team needs a polished click-through UI right now, that's the real constraint.

### Is migrating off NetSuite financially worth it?
Math: NetSuite license $30K to $300K per year, plus $50K to $500K implementation already paid, plus 1 to 2 FTE admins at $120K each. ERPClaw: $0 software, modest hosting cost, optional paid services. Even with conservative assumptions, the payback period is under 12 months for almost every mid-market team. The architectural advantages (open source, AI-native, no vendor lock-in) are upside on top of the cash math.

Related: read the NetSuite comparison, the AI-native ERP framework, the long-form NetSuite alternative post, the migration hub, or the 13-vendor matrix.

---

## Migrate from Odoo to ERPClaw: Export, Map, Validate, Cut Over
URL: https://www.erpclaw.ai/migrate/from-odoo/

# Migrate from Odoo to ERPClaw
SMB playbook. AI-assisted data transcription, schema mapping, cutover. Replace per-user pricing with $0 forever.

## Why teams move from Odoo
Odoo by Odoo S.A. is a capable open-core ERP, but the AI story and the pricing model both create friction. Odoo Enterprise tier-gates the AI features (OdooBot relies on ChatGPT externally). Odoo Community has no AI assistant at all. The Enterprise tier costs $24.90 per user per month base, plus per-app costs that stack quickly for teams using 5 or more advanced modules. Per-user pricing scales with your team and becomes painful past 20 to 30 users.
Pricing basis used on this page: the published tier captured May 24, 2026. Treat it as the source basis for this migration scenario, not as a current quote. For the broader fit decision, read the Odoo alternative guide.
ERPClaw is AI-native at the action layer with no Community-vs-Enterprise split, Free & Open Source, self-hosted, and $0 forever.

- ·No per-user subscription. ERPClaw is Free & Open Source.
- ·No AI tier-gate. AI-native architecture is the default.
- ·No Community-vs-Enterprise gap. Every feature ships under Free & Open Source.
- ·Database-agnostic. SQLite default or PostgreSQL via PyPika.

## What ERPClaw replaces
Module-to-module mapping. Most source functional areas have a clean ERPClaw equivalent.
Odoo
CRM (leads, opportunities, pipeline)

ERPClaw
erpclaw-growth CRM

Odoo
Sales (quotations, orders, invoices)

ERPClaw
erpclaw foundation sales plus erpclaw-integrations-stripe

Odoo
Accounting (chart, journals, taxes)

ERPClaw
erpclaw foundation: the core tables, 12-step GL validation, immutable posted entries

Odoo
Inventory and Warehouse

ERPClaw
erpclaw foundation stock module with immutable lineage

Odoo
Purchase (RFQs, PO, vendor bills)

ERPClaw
erpclaw foundation purchasing module

Odoo
Manufacturing (MRP, BOM, work orders)

ERPClaw
erpclaw-ops manufacturing module

Odoo
Project Management

ERPClaw
erpclaw-ops projects

Odoo
HR (employees, time off, attendance)

ERPClaw
erpclaw foundation HR module

Odoo
Helpdesk

ERPClaw
erpclaw-ops support

Odoo
Field Service

ERPClaw
erpclaw-ops maintenance and erpclaw-fleet

Odoo
Documents and Sign

ERPClaw
erpclaw-documents and erpclaw-esign

## What ERPClaw does NOT replace yet
Honest gaps. If any of these are core to your business, weigh them before committing.

### Odoo POS
Odoo's Point of Sale module is mature with offline support, hardware integrations (cash drawers, barcode scanners, receipt printers), and a polished touch interface. ERPClaw ships erpclaw-pos for the core flows but the hardware integration depth and offline behavior is narrower today. If your retail operation depends on Odoo POS at multiple stores, verify the scope first.

### Odoo Studio (web-based form designer)
Studio lets non-developers add fields, build forms, and define workflows in the browser. ERPClaw takes a spec-first approach: you describe what you want in a SKILL.md spec and an AI agent generates the action. Different model, different audience. If your team relies on Studio's click-to-build UI for daily customization, that workflow shifts.

### Odoo eCommerce module
The eCommerce module is a full website builder plus storefront, blog, and SEO tooling integrated with sales and inventory. ERPClaw integrates with Shopify and Stripe for sales channels rather than running its own storefront. If you depend on the website builder for your public site, plan a separate web stack.

### Large existing Odoo app marketplace
The source ecosystem has a large catalog of third-party modules accumulated since 2005. ERPClaw ships one integrated system across 14 verticals plus the action layer for custom modules generated from spec. Coverage is wide but not yet as deep as that marketplace for very niche needs.

## How the migration works

- 1

### Export from Odoo

The source system exposes structured data two ways. XML-RPC: connect with any client and pull res.partner (customers and vendors), product.product (items), account.account (chart of accounts), account.move (journal entries), sale.order (sales orders), purchase.order (purchase orders). PostgreSQL backup: take a full pg_dump of the database for the safety net. Save each XML-RPC pull as CSV in odoo-export-YYYYMMDD/.

- 2

### Map the schema to ERPClaw

Walk an AI agent through the source to ERPClaw field map. res.partner becomes ERPClaw customer or vendor depending on is_customer / is_supplier flags, product.product becomes item, account.move becomes gl_entry. The agent reads SKILL.md specs for each ERPClaw module and produces a mapping plan before any data moves. Many-to-many relationships need explicit decisions: which tags to keep, which custom fields to port.

- 3

### Install ERPClaw and run the data transfer

Follow /docs/core/install/. SQLite default works for small to mid-sized teams. PostgreSQL is fully supported via PyPika if you already run Postgres. Point the AI agent at the export folder and your fresh ERPClaw install. The agent calls setup-chart-of-accounts, add-account, add-customer, add-vendor, add-item, and post-gl-entries to recreate your books. Studio customizations and custom Python modules need a separate AI conversation per item.

- 4

### Verify and cut over

Run check-gl-integrity in ERPClaw. Compare trial balance, AR aging, AP aging, inventory valuation against the reports you exported. Investigate any drift before going live. Run parallel for one to two weeks for a SMB, then cut over and stop your Enterprise subscription at the next renewal date.

Realistic timeline: 2 to 4 weeks elapsed for a SMB with a standard source install. 4 to 8 weeks for teams using heavy Studio customization and custom Python modules.

## Cost comparison

### Odoo Enterprise

- Base subscription: $24.90 per user per month
- Per-app costs: $7 to $20 per user per month for some advanced apps
- AI assistant: OdooBot uses ChatGPT, Enterprise tier only
- Community edition: Free, no AI assistant, feature-limited

### ERPClaw

- License: $0 forever, Free & Open Source
- Per-app costs: $0, every module free
- AI assistant: AI-native architecture, every action is agent-callable
- No Community-vs-Enterprise split: Every feature ships under Free & Open Source

Pricing reflects the publicly posted Enterprise source used for this page.

## Frequently asked questions

### Can I export my Odoo data?
Yes. XML-RPC supports structured extraction of every model (res.partner, product.product, account.move, sale.order, and the rest). You can also take a full PostgreSQL dump of the database. The AI agent reads the CSV exports and runs ERPClaw import actions; you do not write a migration script by hand.

### What about my Odoo customizations and Studio configs?
Studio customizations and custom Python modules need re-implementation in ERPClaw. The good news: ERPClaw's spec-first regen pattern means an AI agent can generate the new action from a SKILL.md description. Tell the agent what each customization or custom module does and it produces the ERPClaw equivalent. This is typically the variable cost of the migration; budget extra time if you have 20 or more custom modules.

### Will my Odoo Community modules work in ERPClaw?
No direct compatibility. ERPClaw is its own module ecosystem across 14 verticals, not an Odoo fork. If a specific Odoo Community module is core to your workflow, the migration plan needs to either find the ERPClaw equivalent or generate a new ERPClaw module from spec. Most common Odoo Community functionality maps to an existing ERPClaw module.

### How long does it take?
For a SMB with vanilla Odoo (no Studio, few custom modules), expect 2 to 4 weeks elapsed including the parallel run. For a team using heavy Studio customization and custom Python modules, expect 4 to 8 weeks. The AI agent handles bulk data transcription quickly; the customization translation is the variable cost.

### What about my Odoo Enterprise subscription?
Odoo Enterprise is billed monthly or annually per user at $24.90 per user per month (plus per-app costs for some advanced features). Time your migration to cut over at least 30 days before your renewal date so you can stop billing cleanly. Odoo Community is free, so timing is less critical if you are on Community; you just stop using it.

### Both ERPClaw and Odoo are partly open source. Why switch?
Three structural reasons. First, ERPClaw is AI-native at the action layer: every action is callable from a chat agent, with no bolt-on AI layer. Odoo Enterprise's OdooBot relies on ChatGPT externally and is tier-gated to the Enterprise edition. Odoo Community has no AI assistant. Second, ERPClaw is Free &amp; Open Source across the entire codebase with no Community-vs-Enterprise feature split. Third, ERPClaw enforces 12-step GL validation and invariant checks at write time, which Odoo does not.

### Is ERPClaw really free?
Yes. ERPClaw is $0 forever, Free &amp; Open Source, self-hosted. Compare that to Odoo Enterprise at $24.90 per user per month plus per-app costs that can run $7 to $20 per user per month each. For a 10-person team using 5 paid apps on Odoo Enterprise, the annual cost can land between $7,000 and $15,000. ERPClaw drops that to $0 in software cost.

---

## Migrate from QuickBooks to ERPClaw: weekend playbook
URL: https://www.erpclaw.ai/migrate/from-quickbooks/

# Migrate from QuickBooks to ERPClaw
A weekend playbook. AI-assisted transcription, trial balance reconciliation, parallel run, cutover. $0 forever after.

## Why teams move
QuickBooks is fine until it is not. The most common pain points we hear: hit the user cap (BOOM, game over), the data file is laggy and slow, the audit trail has known gaps, and reconciling Stripe + Shopify takes 10 hours per month. ERPClaw is structurally a better fit past the QBO ceiling and free forever.

- ·No user cap. ERPClaw has no per-seat pricing.
- ·Immutable GL. Cancel = reverse, never edit.
- ·First-party Stripe + Shopify. No more manual fee reconciliation.
- ·$0 forever. Self-hosted on your hardware.

## 7-step playbook

- 1

### Export from QuickBooks

QBO: Reports → Trial Balance, Customer List, Vendor List, Item List, Chart of Accounts, Journal entries (full history). Export each as CSV or Excel. QuickBooks Desktop: same reports plus the .qbb backup. Save everything to a folder named qbo-export-YYYYMMDD/.

- 2

### Install ERPClaw on your hardware

Follow the install guide at /docs/core/install/. SQLite default works for most teams. PostgreSQL is supported via PyPika if you already run Postgres. The install completes in 5 minutes plus seed-defaults.

- 3

### Talk to the AI agent

Point an AI agent (Claude Code, Cursor, etc.) at the qbo-export folder and your fresh ERPClaw install. Ask it to import. The agent reads the SKILL.md specs, maps QBO fields to ERPClaw actions (setup-chart-of-accounts, add-customer, add-account, post-gl-entries, etc.), and runs the import.

- 4

### Spot-check the trial balance

Run check-gl-integrity in ERPClaw. Compare your post-import trial balance against the QBO trial balance you exported. They should match to the cent. If they don't, the AI agent helps trace the diff.

- 5

### Spot-check AR and AP aging

Compare the customer-aging and vendor-aging reports between QBO and ERPClaw. Same totals, same buckets. Investigate any drift before going live.

- 6

### Run parallel for one week

Post new transactions in both systems for one business week. Reconcile every day against the prior day's bank deposit and Stripe payouts. By the end of the week, the answer should be obvious: match every day or kick the migration plan back.

- 7

### Cut over

Pick a Monday. Stop posting in QBO. All new transactions land in ERPClaw. Your archived QBO file stays readable for as long as Intuit lets you keep it. The migration is done.

## Common QuickBooks issues, fixed in ERPClaw

### QBO doesn't post Stripe fees natively
ERPClaw's full Stripe integration handles processing fees, application fees, refunds, and ASC 606 revenue recognition. Once you're on ERPClaw, the manual fee reconciliation goes away.

### QBO user cap (5 on Plus, 25 on Advanced)
ERPClaw has no user cap and no per-seat pricing. Self-hosted on your hardware, every user is free.

### QBO data file performance ceiling (laggy and slow at scale)
ERPClaw uses SQLite with WAL mode and indexed queries. Past SQLite's comfortable scale (typically tens of thousands of GL entries), PostgreSQL takes over with no schema change.

### QBO mutable audit trail
ERPClaw enforces immutable GL by constitution. Cancellations create reversal entries; nothing is silently edited. Chain-hashed audit trail and invariant checks make tampering structurally detectable.

### QBO ecosystem lock-in (every connector is a new SaaS subscription)
ERPClaw includes native Stripe and Shopify integrations. Other channels can be ingested via custom actions defined in a SKILL.md spec. One self-hosted system replaces the QBO + connector stack.

## Frequently asked questions

### How long will my QuickBooks-to-ERPClaw migration take?
Most QBO migrations are a weekend of work. Friday: export + install. Saturday: AI-assisted import + trial balance check. Sunday: AR/AP aging spot-check. Monday onward: parallel run for one week. Total elapsed time: about 10 days. Total hours of work: 8 to 12 for most small businesses.

### Will my Stripe sync still break the way it does in QuickBooks?
No. ERPClaw's Stripe integration is first-party with every operation. Processing fees, application fees, refunds, payouts, and ASC 606 revenue recognition are all built in. The 12-step GL validation runs on every transaction. The manual fee reconciliation in QBO disappears.

### What happens to my QuickBooks audit trail history?
Your QBO data stays in QBO until you cancel the subscription. ERPClaw imports the historical journal entries so your year-over-year reports work post-migration. You keep both systems readable for as long as you want; ERPClaw is your write-side from the cutover date forward.

### What if I'm on QuickBooks Desktop instead of QBO?
The same approach works. Export the same reports plus the .qbb backup. The AI agent reads the structured exports. QBD's IIF format is well-known and the agent handles it. Some QBD-specific quirks (like inactive class hierarchies) need a manual decision; the agent flags them.

### Will my chart of accounts come over cleanly?
Yes. The QBO chart of accounts exports to CSV and the AI agent imports each account into ERPClaw with its type (asset, liability, equity, income, expense), parent hierarchy, and active status. ERPClaw's account model maps cleanly to QBO's. You can simplify post-migration if you've accumulated unused accounts.

### Can I still let my CPA review the books in QuickBooks format?
ERPClaw exports trial balance, P&L, balance sheet, AR aging, AP aging, GL detail, and all 20-plus financial reports as CSV or PDF. Your CPA can review in their preferred format. If they specifically want a QBO file format, you keep the archived QBO subscription read-only.

### Is this really $0 after migration?
ERPClaw software cost is $0 forever. You decide your hosting cost: $0 on your own laptop or office server, $5 to $20 per month for a small VPS, more if you want a managed cloud setup. No software subscription. No per-seat fee. No metered usage. The only ongoing cost is whatever you choose for hosting infrastructure.

Related: read the QuickBooks comparison, the AI accounting story, the long-form QuickBooks alternative post, the migration hub, or the install guide.

---

## Migrate from SAP Business One to ERPClaw
URL: https://www.erpclaw.ai/migrate/from-sap-business-one/

# Migrate from SAP Business One to ERPClaw
Mid-market playbook. AI-assisted data transcription, schema mapping, verification, cutover. Replace per-user maintenance with $0 forever.

## Why teams move from SAP Business One
SAP Business One by SAP SE is a mature mid-market ERP, but the economics are tough on a growing team. License is $1,500 to $3,000 per user one-time (list price), maintenance is roughly $94 to $148 per user per month, and partner-led implementation typically runs $30,000 to $200,000. The AI story is a separate gap: SAP Joule is shipped for S/4HANA, not Business One, so B1 customers cannot access SAP's flagship AI assistant today. Add in vendor lock-in to the SAP ecosystem and the recurring maintenance bill, and the migration math gets compelling.
ERPClaw is AI-native from the action layer up, Free & Open Source, self-hosted, and $0 forever. No per-user fee, no partner contract, no AI tier-gate.

- ·No per-user maintenance. ERPClaw is Free & Open Source.
- ·AI-native, not retrofitted. Every action callable from a chat agent.
- ·No partner dependency. Most teams self-migrate with an AI agent.
- ·Database-agnostic. SQLite default or PostgreSQL via PyPika.

## What ERPClaw replaces
Module-to-module mapping. Most SAP Business One functional areas have a clean ERPClaw equivalent.
SAP Business One
Financials, Chart of Accounts, Journal Entries

ERPClaw
erpclaw foundation: the core tables, 12-step GL validation, immutable posted entries

SAP Business One
Sales (AR Invoices, Sales Orders, Deliveries)

ERPClaw
erpclaw foundation sales module plus erpclaw-integrations-stripe for online sales

SAP Business One
Purchasing (AP Invoices, Purchase Orders, Goods Receipts)

ERPClaw
erpclaw foundation purchasing module

SAP Business One
Inventory and Warehouse Management

ERPClaw
erpclaw foundation stock module plus stock_ledger_entry with immutable lineage

SAP Business One
Business Partners (Customers, Vendors, Leads)

ERPClaw
erpclaw foundation party module plus erpclaw-growth CRM

SAP Business One
Production (basic manufacturing)

ERPClaw
erpclaw-ops manufacturing module

SAP Business One
MRP

ERPClaw
erpclaw-ops planning

SAP Business One
Service Management

ERPClaw
erpclaw-ops support

SAP Business One
Project Management

ERPClaw
erpclaw-ops projects

SAP Business One
Fixed Assets

ERPClaw
erpclaw-ops assets

SAP Business One
Banking and Payments

ERPClaw
erpclaw foundation payments plus erpclaw-integrations-stripe

## What ERPClaw does NOT replace yet
Honest gaps. If any of these are core to your business, weigh them before committing.

### HANA-backed real-time reporting
SAP HANA's in-memory column store powers very fast pivot reporting on large datasets. ERPClaw on PostgreSQL handles tens of millions of GL entries cleanly, but a HANA tenant tuned for sub-second cross-cube reporting on a billion rows is in a different class. If your finance team relies on HANA-specific analytics, plan a separate reporting layer.

### Global partner network
SAP has a worldwide partner network with implementation specialists in most countries. ERPClaw is community-supported plus paid services from AvanSaber Inc. If you are in a region where local SAP partners are the only ERP vendor option, that is a real consideration.

### Mature multi-jurisdictional payroll for certain EU and APAC countries
SAP Business One localizations cover detailed payroll rules in many EU and APAC jurisdictions. ERPClaw ships erpclaw-regions for US, UK, CA, EU, and IN core tax and reporting, but country-specific payroll edge cases need verification before you assume parity.

### Some industry-specific add-ons
SAP Business One has a large catalogue of partner-built industry add-ons accumulated over twenty years. ERPClaw ships 14 native verticals (healthcare, education, construction, food, hospitality, retail, automotive, agriculture, property, legal, nonprofit) plus an action layer for custom verticals. The coverage is wide but not yet as deep as the SAP partner catalogue for very niche industries.

## How the migration works

- 1

### Export from SAP Business One

SAP Business One stores data in Microsoft SQL Server or SAP HANA. Use the SAP B1 SDK or a direct SQL backup to extract: Chart of Accounts (OACT), Business Partners (OCRD), Items (OITM), Journal Entries (OJDT + JDT1), AR Invoices (OINV), AP Invoices (OPCH), Inventory (OITW), and User-Defined Fields. Save each as CSV in a folder named sap-b1-export-YYYYMMDD/. Keep a full backup of the underlying database as a safety net.

- 2

### Map the schema to ERPClaw

Walk an AI agent through the SAP Business One field map. OACT.AcctCode becomes ERPClaw account_code, OCRD.CardCode becomes customer or vendor id, OITM.ItemCode becomes item code, OJDT lines become gl_entry rows. The agent reads SKILL.md specs for each ERPClaw module and produces the mapping plan before any data moves.

- 3

### Install ERPClaw and run the data transfer

Follow /docs/core/install/. PostgreSQL is recommended for SAP Business One migrations because the historical data volume is usually significant. Same code targets SQLite or PostgreSQL via PyPika. Then point the AI agent at the export folder and your fresh ERPClaw install. The agent calls setup-chart-of-accounts, add-account, add-customer, add-vendor, add-item, and post-gl-entries to recreate your books. Localizations, user-defined fields, and stored procedures need a separate AI conversation per item.

- 4

### Verify and cut over

Run check-gl-integrity in ERPClaw. Compare trial balance, AR aging, AP aging, and stock valuation against the SAP Business One reports you exported. Investigate any drift before going live. Run parallel for two to four weeks for a 20 to 50 person team, then cut over and notify your SAP partner that you will not renew the annual maintenance.

Realistic timeline: 4 to 8 weeks elapsed for a 20 to 50 person team with vanilla SAP Business One. 8 to 12 weeks for heavily customized tenants with extensive user-defined fields, stored procedures, and partner add-ons.

## Cost comparison

### SAP Business One

- License: $1,500 to $3,000 per user, one-time
- Maintenance: $94 to $148 per user per month, ongoing
- Partner implementation: $30,000 to $200,000 one-time
- AI assistant: Not available on B1 (Joule is S/4HANA only)

### ERPClaw

- License: $0 forever, Free & Open Source
- Maintenance: $0 ongoing
- Implementation: $0 self-migration with AI agent, optional paid services from AvanSaber Inc.
- AI assistant: AI-native architecture, every action is agent-callable

SAP Business One pricing reflects publicly reported list prices and partner quotes. Actual pricing depends on your partner contract.

## Frequently asked questions

### Can I export my SAP Business One data?
Yes. SAP Business One stores data in Microsoft SQL Server or SAP HANA, both of which support full SQL export. The SAP B1 SDK also exposes a DI API for structured extraction. The AI agent reads the CSV exports and runs ERPClaw import actions; you do not write a migration script by hand.

### How long does the migration take?
For a 20 to 50 person team with vanilla SAP Business One, expect 4 to 8 weeks elapsed including the parallel run. Heavily customized B1 tenants with extensive user-defined fields, stored procedures, and partner add-ons take 8 to 12 weeks. The AI agent handles bulk data transcription quickly; the customization translation is the variable cost.

### What happens to my existing SAP partner contract?
SAP Business One is typically sold through a partner with an annual maintenance contract covering the license and support. Read your contract carefully. Most are annual with a renewal date; you give notice before that date and stop paying. The license itself was a one-time purchase ($1,500 to $3,000 per user list price) and is yours to keep, but you lose access to maintenance updates and support. ERPClaw migrations should be timed so the cutover lands at least 60 days before your SAP renewal.

### Will ERPClaw handle our multi-entity consolidation?
Yes for most US, UK, and India multi-entity setups. ERPClaw supports multiple companies in a single shared database with per-entity GL, multi-currency at 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED), and entity-level reporting. Where ERPClaw is honestly different from SAP Business One is transactional FX revaluation at scale, which is intentionally out of ERPClaw's scope, not a roadmap item, and automatic consolidation-time translation, which is recorded manually today. If your group has 10+ entities with daily FX revaluation needs, verify the scope before committing.

### What if we need SAP-specific features ERPClaw doesn't have?
Be specific about which feature. The most common SAP-only items we hear are HANA analytics, certain country payroll localizations, and very niche industry add-ons. ERPClaw covers the core ERP primitives (financials, sales, purchasing, inventory, manufacturing, projects, assets) and 14 verticals; if your gap is in an area we ship, the AI agent re-expresses your SAP customization as an ERPClaw action. If the gap is in an area we have not yet shipped, the spec-first regen pattern means a new action can be generated from a SKILL.md description.

### Is ERPClaw really free given the SAP license cost we already paid?
Yes. ERPClaw is $0 forever, Free &amp; Open Source, self-hosted. The sunk cost of your existing SAP license does not change the forward math: SAP maintenance is roughly $94 to $148 per user per month ongoing, plus any partner support retainer. Switching to ERPClaw drops that to $0 in software cost. Your only forward cost is hosting infrastructure (which you may already run for SAP B1).

### Why move from SAP Business One to ERPClaw?
Three structural reasons. First, ERPClaw is AI-native: every action is callable from a chat agent, with no bolt-on AI layer. SAP Joule is the AI assistant for S/4HANA, not Business One, so B1 customers cannot access SAP's main AI offering today. Second, ERPClaw is Free &amp; Open Source with no per-user fees, replacing a $94 to $148 per user per month maintenance bill. Third, ERPClaw avoids the partner-dependent implementation model: most teams self-migrate with an AI agent in 4 to 8 weeks instead of a 6 to 18 month partner engagement.

---

## Open-Source AI Accounting, AI-Native, Free
URL: https://www.erpclaw.ai/open-source-ai-accounting/

# Open-source AI accounting: ERPNext is bolt-on. ERPClaw is AI-native.
ERPNext is the open-source default but its AI is a plug-in stack: changAI, NextAI, Ollama integrations, Composio. Per ChatGPT's own framing. ERPClaw is the first open-source AI-native ERP we've found. Free & Open Source, self-hosted, free forever, full ERP scope. Here's the honest comparison.
covers every major business function·Free & Open Source·self-hosted·$0 forever
Install in 4 commands Read the source
By Nikhil Jathar, Co-founder, ERPClaw

Looking for the broader AI accounting overview (not open-source-specific)? See our main AI accounting page.

## What is open-source AI accounting?
Open-source AI accounting is accounting software where an AI agent runs your books and the source code is public under an OSI-approved license you can read, fork, and self-host. ERPNext is the open-source default, but its AI is a plug-in stack (changAI, NextAI, Ollama, Composio) on top of a forms-and-workflows ERP. ERPClaw is the only AI-native open-source option built around the agent from the floor up.

- ·Open source license: ERPClaw is Free & Open Source; ERPNext is Free & Open Source; Akaunting is open source; Manager.io is closed-source freeware.
- ·AI-native vs AI-decorated: ERPClaw is AI-native by design; ERPNext, Akaunting, Odoo Community are AI-decorated bolt-ons.
- ·Self-hosted: all of these can self-host. ERPClaw runs on SQLite by default; PostgreSQL is fully supported via PyPika.
- ·Scope: ERPClaw is full ERP covering every major business function across 14 industry verticals; most open-source AI accounting tools are accounting-only.
- ·Cost: ERPClaw is $0 forever. No subscription, no per-seat pricing.

Open up the search results for open-source AI accounting and the same names come back: ERPNext, Akaunting, Odoo Community, Manager.io. ERPNext is the default. Ask ChatGPT what AI it actually has and you get a straight answer.
"ERPNext AI is mostly add-ons: changAI (AI agent for natural-language queries), NextAI (Frappe app using Gemini), Ollama/ChatGPT (custom integration), Composio (API layer)."
Translation: the AI is a plug-in stack glued onto a forms-and-workflows ERP. That is not the same as being built around AI from the floor up. This page is the honest map of what open-source AI accounting looks like in 2026, what AI-native actually means in this category, and where ERPClaw fits as the open-source option built AI-native by design with an open source license, self-hosted, free forever.
For the broader category framework (AI-decorated vs AI-native across all ERP, not just open-source), read our pillar on AI-native ERP. This page is the open-source slice. The audience here is the technical buyer who has already rejected proprietary SaaS on cost, control, or lock-in grounds, and now wants the real answer to which open-source option is built around AI from the floor up rather than around a forms-and-workflows ERP with AI bolted on after the fact.
Quick orientation on ERPClaw before we get into the comparison: full ERP scope across every major business function, Free & Open Source, self-hosted, free forever, patent pending, trademark filed. SQLite is the default database; PostgreSQL is fully supported via PyPika. The Stripe Marketplace listing is live and selling. The Shopify integration is shipped at v1.1.3, though the Shopify App Store listing itself is still pending Partners review. The rest of this page is the architectural argument behind that scope.

## The 5-trait test for AI-native (vs AI-decorated)
Before naming any vendor, here is a testable framework. Take any tool, score it on these five traits, and the AI-decorated ones fall out fast. The framework comes from a ChatGPT GPT-5 conversation we captured on 2026-05-05; we've added the ERPClaw column on the right.
| Trait | AI-decorated | AI-native | ERPClaw concrete
| User experience | Forms, menus, dashboards. AI lives in a sidebar. | Conversation is the interface. Forms are an artifact. | Type "set up a company called Acme Imports" and the AI invokes setup-company. 94 GAAP accounts seed in one transaction.
| Workflow | Hard-coded business rules. Each new flow needs a developer. | Reasoning over the action catalog. New flows compose from existing actions. | "Add Bob from BigCo as a customer with net-30 terms" maps to add-customer with one DB write. No new code path needed.
| Data | Tables and forms. Context is the user's job. | Every action accepts context (descriptions, prior history, vendor patterns) and resolves to structured writes. | One shared SQLite or Postgres backend, full schema visible to the agent at request time.
| Automation | Scripts. Rules engines. Cron jobs. | Agents that pick the next action from a goal. Same primitives the human uses. | Every action across the platform is invokable from natural language. No separate "agent API" layer.
| Governance | Roles + audit logs sit beside the AI. The AI can bypass them. | AI controls run inside the action. Confirmation flags, GL invariants, and audit trails are part of the write path. | Every write action requires a per-invocation `--user-confirmed` flag. Every action writes one row to immutable `audit_log` with before/after JSON.

Trait taxonomy: ChatGPT GPT-5, 2026-05-05. ERPClaw column: our own architecture.
The pattern across the five traits is the same one. AI-decorated tools keep the human as the primary user and let the AI suggest. AI-native tools promote the AI to primary user and put the controls (confirmation flags, invariants, audit) inside the action itself. For a longer treatment of how this plays out on the books side, read our piece on open-source AI accounting and what makes it work end to end.

## How the open-source AI accounting tools compare
Five tools that come up most often when people search for open-source AI accounting. License, AI architecture, hosting, database, action breadth, plus the honest strength and gap for each. ERPClaw on the bottom row.
| Tool | License | AI architecture | Hosting | DB backend | Action breadth | Key strength | Key gap
| ERPNext | Free &amp; Open Source | Add-ons (changAI, NextAI, Ollama) | Self-host or Frappe Cloud | MariaDB | Full ERP | Mature, multilingual, 13-year history | AI is bolt-on; some AI features Enterprise-framed
| Odoo Community | LGPL v3 | AI features Enterprise-only | Self-host | PostgreSQL | Full ERP | Huge addon ecosystem | AI behind paywall in Community edition
| Akaunting | Free &amp; Open Source | No AI | Self-host or Akaunting Cloud | MySQL | Bookkeeping only | Simple SMB bookkeeping | Not AI-native; not full ERP
| Manager.io | Proprietary (free desktop) | No AI | Desktop / self-host | Proprietary | Bookkeeping only | Strong basics, zero install friction | Not open source; no AI
| ERPClaw | Free &amp; Open Source | AI-native (action layer is the API) | Self-host (CLI / Docker) | SQLite default; PostgreSQL fully supported | Full ERP across every major business function | AI-native by design + open + self-host + free | CLI-first today; web UI in progress

ERPNext has the largest install base in this group and the most multilingual depth. 13-year history, mature regional accounting, real customers across continents. The AI is the gap. ChatGPT's own framing is that ERPNext AI is "mostly add-ons": changAI for natural-language queries, NextAI as a Frappe app calling Gemini, Ollama and ChatGPT custom integrations, Composio as the API glue layer. None of that is wrong, none of it is bad, and ERPNext is a real piece of software. It just isn't AI-native; the AI sits on top of an ERP designed before AI mattered.
Akaunting and Manager.io are bookkeeping tools, not full ERP, and neither markets an AI agent as a core feature today. Akaunting is Free & Open Source with a SaaS option; Manager.io is proprietary but free for desktop. Both do the basics well (chart of accounts, invoicing, expenses, basic reports) and stop there. If your need is small-business bookkeeping with a clean UI and you don't need AI or a full ERP backbone, they are honest options. If you came to this page searching for open-source AI accounting, the AI half of the pitch is the gap.
Odoo Community has the scope. LGPL v3, self-host on Postgres, large addon ecosystem covering manufacturing, CRM, inventory, e-commerce. The catch is that most of the AI features in the Odoo brochure ship in Odoo Enterprise, not in the Community edition. ERPClaw doesn't gate. Every action across every module is in the open Free & Open Source repo. If you compare AI-native peers, the closest one in spirit is the AI-native cloud option Rillet, which is closed-source and SaaS-only. ERPClaw is the open-source-and-self-hosted version of that architecture choice.

## Hosting, license, database: the practical matrix
The technical-buyer question underneath the AI question is always the same. Can I deploy this where I need to? Three quick answers.

### License: Free & Open Source (open source)
ERPClaw ships under Free & Open Source (effective v4.2.0; v4.1.x and earlier were MIT). ERPNext and Akaunting are Free & Open Source. Odoo Community is LGPL v3. All four are copyleft, auditable, and self-hostable; the license is a checkbox the buyer reviews once. The wedge between ERPClaw and the others is architecture, not license: ERPClaw is AI-native by design and ERPNext is forms-and-workflows with AI plug-ins. Read the LICENSE file in the repo, audit the GL math, ship it in your stack.

### Hosting: anywhere SQLite or Postgres runs
ERPClaw is self-hosted. Your laptop, your VPS, your AWS Lightsail box, your Docker host, your air-gapped server. The data lives at `~/.openclaw/erpclaw/data.sqlite` by default. There's no cloud dependency, no phone-home, no per-seat license server. We run a test server on AWS Lightsail Ubuntu 24.04; that's the reference deployment, not a requirement. If you can ssh into it and run Python 3.10, ERPClaw runs on it.

### Database: SQLite default, Postgres first-class
ERPClaw is database-agnostic via PyPika, the SQL builder layer. SQLite is the default for zero-setup, file-based simplicity. PostgreSQL is fully supported as an alternative backend, same code, same actions, same test suite. This is not a fallback; Postgres is a first-class option for multi-user and production scale. Switch by setting an env var, no rewrite. If your prior assumption was "open-source means SQLite-only", drop that. Both are real here, and the same 12-step GL validation path runs against either backend.

## What ChatGPT itself says
We're not the only ones drawing this line. When you ask GPT-5 to score open-source ERPs on the AI-native vs AI-decorated axis directly, it splits the same way. Here's the verbatim summary it produced for the ERPNext side, with attribution:
"ERPNext AI is mostly add-ons: changAI (AI agent for natural-language queries), NextAI (Frappe app using Gemini), Ollama/ChatGPT (custom integration), Composio (API layer)."
Same conversation, on Odoo Community: many of the AI features highlighted in marketing live in Odoo Enterprise, not the Community edition. So the open-source path forward, if AI-native is a hard requirement, is narrow. ERPNext (AI as add-ons), Odoo Community (AI behind paywall), Akaunting and Manager.io (bookkeeping only), or ERPClaw (AI-native, open source, free).
That's the honest field. Pick the one whose tradeoffs match your project, and don't let any vendor (us included) tell you the choice is obvious.

## What ERPClaw doesn't do (yet)
Trust is built by naming the gaps before the buyer finds them. Here are four real ones, as of mid-2026.

### Web UI for Stripe and Shopify is still CLI-first
webclaw chat-and-CLI is the primary interface. We have an Astro dashboard for marketing and docs, but the configuration UI for Stripe and Shopify lives in the CLI today. We're shipping web settings panels next, but the AI is the primary user, so chat works for everyone.

### Shopify App Store listing is pending
Stripe Marketplace listing is live and selling. The Shopify equivalent is pending Partners-dashboard review. The Shopify integration itself is shipped at v1.1.3 with OAuth pairing, status mirror, GDPR webhooks, and a daemon. You can install from GitHub today.

### Deepest in the US; English-only docs
Tax depth is US: W-2, 1099, NACHA, FICA, FUTA, SUTA. We have 4 regional modules (CA, EU, IN, UK) but English docs only. ERPNext has more multilingual depth and more mature regional payroll outside the US. If you're billing in 5 languages, ERPNext is more battle-tested today.

### Multi-currency Phase 1 only
ERPClaw supports 7 currencies (USD, EUR, GBP, CAD, INR, SGD, AED). Invoice currency must equal payment currency. There's no FX gain/loss in our books today. Stripe handles the conversion if your customer pays from a different country, and our books stay clean.

## Who picks open-source AI accounting
Four profiles where open-source AI accounting is the default-correct answer, not the contrarian one.

### Founders and CTOs of small businesses
You read the code before you trust the tool. You don't want SaaS bills that climb with every seat. You want the AI to actually post journal entries, not just read your dashboard. AI accounting + AI bookkeeping cover the founder flow.

### Mid-market finance leaders evaluating off-cloud
QuickBooks doesn't scale. NetSuite is six figures. You want the primitives without the lock-in. Compare against NetSuite, Sage Intacct, and QuickBooks before you sign a renewal.

### ERPNext / Odoo evaluators
You like open source but want AI-native, not bolt-on. ERPNext's AI is plug-ins (changAI / NextAI / Composio); Odoo's AI features are Enterprise-gated. The side-by-side detail: ERPClaw vs ERPNext and vs Odoo.

### CPAs and audit-aware teams
You want a system whose controls you can read end to end. Immutable GL, chain-hashed audit trail, full invariant pipeline. Free & Open Source source on your hardware. The same defensibility model that puts Linux on production servers, applied to the books.

## Pricing
$0 forever
Free & Open Source license. Self-hosted. Unlimited users. No per-transaction fee. No upgrade tier. No "AI add-on" subscription.
| Tool | License | Indicative pricing
| ERPClaw | Free & Open Source (open source) | $0 forever, every module included
| ERPNext (Frappe Cloud) | Free & Open Source + paid managed cloud | Frappe Cloud from around $25/site/month per founder-reported quotes; self-host is $0
| Odoo Community | LGPLv3 (Community); proprietary Enterprise | Community $0; Enterprise from around $25 to $59 per user/month plus apps
| Akaunting | GPLv3 + paid SaaS | Self-host $0; SaaS from a few dollars per month
| Manager.io | Freeware (closed source) + Cloud | Desktop free; Cloud and Server tiers paid

Pricing is the lever that compounds. QuickBooks Plus is $85/month (about $5,100 over five years). NetSuite + implementation can reach six figures. Open source removes the lever entirely. See our pricing page for the full breakdown.

## Getting started in 4 commands
Here is the complete command-line install path from ClawHub to your first company.
`# Install ClawHub if you haven't
pip install clawhub

# Install ERPClaw foundation
clawhub install erpclaw

# Start the chat agent
openclaw chat

# Now talk to it: "set up a company called Acme Imports"`Those 4 commands install the foundation (the action layer plus the module registry that knows about every available module). When you describe your industry to the chat agent ("I run a Shopify store", "I run a school", "I run a dental practice"), it auto-pulls the matching vertical module from GitHub and seeds the right tables. No second install command, ever. The database lands at `~/.openclaw/erpclaw/data.sqlite` by default. Want PostgreSQL instead? Set `ERPCLAW_DB_PATH=postgres://...` before the install and the same code path runs on Postgres. More on what the foundation does at features/foundation.
The chat agent is the primary interface today. If you prefer the raw action API, every action is callable directly via `--action` on the relevant `db_query.py` script for the module that owns it. Output is JSON to stdout. Errors are JSON with a non-zero exit code. So you can wire ERPClaw into any pipeline (cron, CI, an upstream agent of your own) without going through the chat layer at all.
Full install docs: /docs/core/install/. Source: github.com/avansaber/erpclaw.

## Frequently asked questions

### Is ERPClaw really open source?
Yes. Every line is Free & Open Source. The avansaber/erpclaw repo is public, no Enterprise-only features, no paid upsell tier. Every action across the platform is in the open repo. Read the source, fork it, run it on your laptop, ship it inside a client deployment.

### What's the difference between AI-native and AI-decorated?
AI-decorated means a forms-and-workflows ERP with a chatbot or copilot bolted on. The AI suggests, a human clicks a button, the underlying form runs. AI-native means the action layer is the API. Every action is invokable directly from a natural-language prompt. The AI doesn't translate then form-fill, it submits the action with structured arguments and full GL validation.

### Can ERPClaw replace QuickBooks for a small business?
For US SMBs with USD invoicing and simple payroll, yes. ERPClaw covers AR, AP, GL, payroll, taxes, inventory, and Stripe and Shopify integration end to end. Bank coverage runs both ways: a Plaid connector links your accounts and syncs transactions, and the statement importer reads OFX, CAMT.053, MT940, and BAI2 files with auto-matching against open invoices and payments. Where it is genuinely behind QuickBooks is the polished consumer-grade web experience, not the bank connection.

### Does ERPClaw work with PostgreSQL or only SQLite?
Both are first-class. ERPClaw is database-agnostic via the PyPika SQL builder. SQLite is the default for zero-setup, file-based simplicity. PostgreSQL is fully supported for multi-user and production scale. Same code path, same actions, same test suite. Switch backend by setting an env var; no rewrite required.

### Is ERPClaw free forever, or is there a paid tier?
Free forever. Open source under Free & Open Source. No paid tier, no Enterprise edition, no per-seat pricing. The full action catalog across every module is open to everyone. We make money on optional services (deployment help, custom integrations, training); the software itself stays free.

### How does ERPClaw compare to ERPNext?
Both are open-source full ERP. ERPNext has a 13-year history, multilingual support, and mature regional payroll. Its AI is an add-on stack: changAI, NextAI, Ollama integrations, Composio. ERPClaw is younger but AI-native by design. The action layer is the API. If multilingual depth matters most, pick ERPNext. If AI-native architecture matters most, pick ERPClaw.

### Can I migrate from QuickBooks Online to ERPClaw?
Yes, with manual import today. ERPClaw has CSV import for chart of accounts, customers, vendors, items, and opening balances. Stripe and Shopify connectors handle live transaction history once connected. We don't have a one-click QBO migrator yet, the data flow is CSV export from QBO, mapping pass, then CSV import into ERPClaw. Doable in an afternoon for a typical SMB.

For the deeper category framing across all ERP (not just open-source), read /ai-native-erp/.

## Where to go next

### Try it now
`clawhub install erpclaw` and you're up in under 60 seconds. Full install docs link inside.

### Read the category piece
The framework for evaluating any AI-native ERP, open-source or otherwise. Useful before you talk to any vendor.

### Talk to a co-founder
Ask us anything through the support page. Real questions, real founder answers, no sales layer in between. Pricing is $0; the call is too.

## Sources

- ChatGPT GPT-5 conversation captured 2026-05-05. Two verbatim quotes used in this article: ERPNext AI as add-ons (changAI / NextAI / Ollama / Composio), and Odoo Community AI features being Enterprise-gated. Captured to dated copy on file.
- ERPClaw `module_registry.json`: signed, version-pinned catalog of every module and action. Source: github.com/avansaber/erpclaw/scripts/module_registry.json.
- ERPClaw `LICENSE` file (Free & Open Source, 2026; v4.1.x and earlier MIT). Source: github.com/avansaber/erpclaw/LICENSE.
- Stripe Marketplace listing: marketplace.stripe.com/apps/erpclaw-accounting. Live, with real production users.
- Comparison table cross-checks: ERPNext (frappe.io), Odoo (odoo.com/page/community), Akaunting (akaunting.com), Manager.io (manager.io). License and feature claims verified against each project's public documentation.
Author: Nikhil Jathar, Co-founder, ERPClaw. Last updated 2026-05-05.

---

## Pricing - Free Open Source ERP, Self-Host for $0
URL: https://www.erpclaw.ai/pricing/

# Simple, transparent pricing
ERPClaw is 100% Free & Open Source: the whole ERP, self-hosted on your hardware, $0 forever. Not freemium, not time-limited. Your data never leaves your server.

Self-Hosted
$0/forever
The full ERP on your own infrastructure. Your data, your server, your rules. No strings attached.

- Every module included
- Every operation available from day one
- Unlimited users
- Unlimited companies
- Full source code (open source license)
- Community support via GitHub
- Database agnostic, PostgreSQL or SQLite: your choice, same software
- ERPClaw OS, self-improving engine included
- Works on any messaging platform
- Web dashboard included (webclaw)
- 14 industry verticals
- 4 regional compliance suites
Download from GitHub
By request
Cloud Managed
Priced at onboarding
Everything in Self-Hosted, plus we handle hosting, updates, backups, and support so you can focus on running your business.

- Everything in Self-Hosted
- Fully managed hosting
- PostgreSQL managed database
- Automatic updates and patches
- Priority email support
- Daily encrypted backups
- 99.9% uptime SLA
- Monitoring and alerting
- Onboarding assistance

"Every ERP pricing page is an opening offer in a negotiation. Ours is a statement: zero dollars, forever, for everyone, because free and open source is the business model, not a trial of it. The only thing we will ever charge for is running it so you don't have to."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## Detailed comparison
Both plans include the complete ERP. The difference is who manages the infrastructure.

Feature
Self-Hosted
Free

Cloud Managed
By request

Skills / Modules
All included
All included

Users
Unlimited
Unlimited

Companies
Unlimited
Unlimited

Accounting & GL

Inventory & Manufacturing

HR & Payroll

CRM & Helpdesk

AI Engine & Analytics

Billing & Subscriptions

Source Code Access

Hosting
You manage
We manage

Updates
You run one update command
Automatic

Backups
You copy one file
Daily, encrypted

Support
Community (GitHub)
Priority email

Uptime SLA
N/A
99.9%

Onboarding Help
Documentation
Guided setup

The price is simple because the model is simple: the software is free, and you choose who hosts it. See AI-native ERP for why ERPClaw can charge $0 forever, or foundation features for what's actually in the box. Comparison versus the paid alternatives: NetSuite, Rillet, Odoo.

## Frequently asked questions

Is ERPClaw really free? Yes. ERPClaw is 100% open source as Free &amp; Open Source software. Every module, every operation, and the ERPClaw OS self-improving engine are available on GitHub at no cost. There are no hidden features, no premium tiers gated behind a paywall, and no usage limits. Self-host it on your own server and use it forever without paying a cent. The Cloud Managed plan is an optional convenience for teams who prefer not to handle server administration.
What is actually included? Everything. The books (double-entry accounting, payments, bank reconciliation, US tax), the operations (inventory, purchasing, manufacturing), the people side (built-in Sales, HR, payroll), and the industry editions, all in one install. There is no feature matrix to study because there are no tiers: every module is included, free.
Can I start self-hosted and migrate to Cloud later? Absolutely. Your entire ERPClaw instance runs on a single database (SQLite by default, PostgreSQL optional). Migrating to Cloud Managed is as simple as uploading your data to our managed PostgreSQL infrastructure. All your data, including chart of accounts, transactions, GL entries, customers, inventory, and employees, transfers intact. No data transformation, no import wizards, no downtime beyond the few minutes it takes to migrate.
Do you offer enterprise support or custom deployments? Yes, by request. For dedicated support, custom integrations, or an on-premises deployment with a support contract, reach out to us at support@avansaber.com and we will scope it with you. The GitHub community and documentation cover most day-to-day questions.

---

## ERPClaw: the Free, Self-Hosted AI-Native ERP
URL: https://www.erpclaw.ai/product/

# The AI-native ERP. Run your whole business in plain language.
Free forever, self-hosted, PostgreSQL or SQLite. You say what you want done. Deterministic accounting code does it, and an invariant engine checks the books after every posting. The AI decides what you meant. It never decides what the books say.
Get ERPClaw on GitHubRead the install guide

## What "in plain language" actually means
Not a chatbot bolted onto menus. The request is the interface, and the result is a posting in a double-entry ledger.
"Invoice Acme for the March retainer, 4,500 dollars, due in 30 days."
A real invoice posts to a real ledger, numbered, dated, and aged from today.

"What did we actually make on the Henderson job?"
Revenue minus every cost tagged to that job, from the books, not from a spreadsheet copy.

"Hold 40 of the blue chairs for this order."
A hard reservation nobody else can quietly consume, refused up front if the stock is not there.

"Import the January bank statement and match it to open invoices."
Statement lines land, match themselves where they can, and queue the rest for your call.

"Bill this client for exactly the usage they racked up this month."
Usage rated, invoices generated, books posted: any pricing shape you can describe in a sentence.

"Show me everything we know about this customer."
One search across their deals, invoices, payments, and history, because it is all one system.

## One system of record, not a stack of slices
Every major business function, one install, one database, one place the truth lives. Nothing is a paid add-on, because nothing here is paid.

### Money
Accounting, invoicing, bills, payments, bank statement import, multi-entity consolidation, and reports that split by department, project, or any label you run the business on.

### Things
Inventory with hard reservations, putaway routing, guided pick lists, item substitutes, manufacturing with work orders and subcontracting, and honest landed costs.

### People and deals
Built-in Sales from first lead to closed order: contacts, follow-up tasks, custom pipeline stages, saved views, and search across every sales record.

### The awkward rest
Purchasing, billing any pricing shape you can describe, tax, HR, payroll, support, and the industry editions that install when you say what your business is.

See the full feature tour

## PostgreSQL or SQLite. Your call, same code.
SQLite is the zero-setup default, which is why the install takes minutes. PostgreSQL is a first-class backend for teams that already run it, not a fallback and not an upgrade fee. The same code runs on either, and switching is a configuration choice rather than a migration project.

## Free forever is the architecture, not a promo
Open source under Free & Open Source, self-hosted on your own machine, with your books in a database you can open yourself. There is no free tier that expires, no per-seat meter, and no export ransom, because the business model is not your data. Paid managed hosting, where we run it for you, is available by request through the pricing page; the software itself stays $0.

Product directory

## Choose how to explore ERPClaw
Start with the complete product, add read-only Insights, review the operating layer, compare deployment choices, or follow an order through the business.
ERPClawThe complete AI-native ERPERPClaw InsightsRead-only answers across your businessERPClaw OSThe operating layer behind the productPricingSoftware cost and deployment choicesChat demoFollow a guided business conversationOrder walkthroughFollow one order from stock to books

## Not ready to move your books?

### Start with ERPClaw Insights
Your AI analyst. Connect Stripe or Shopify and get plain-language answers about cash, customers and costs. No migration, nothing to replace.

### See it speak your industry
Editions for retail, manufacturing, and schools that set themselves up when you say what your business is.

---

## Quality: Tested in Layers, Books Balanced to the Cent
URL: https://www.erpclaw.ai/quality/

# Six layers of tests. Six gates on every release.
Books never broke.
We don't ship hope. We ship proof. Layered automated tests, all passing on every release. GL balanced to the cent across $1,060,433.00 in transactions.

erpclaw-test-suite, run on every change
PASSCreate invoice Wayne Enterprises $12,5000.3s
PASSSubmit - GL entries posted, balanced0.2s
PASSCancel invoice - 2 reversal entries created0.1s
PASSTrial balance: $1,060,433 DR = $1,060,433 CR0.4s
BLOCKSQL injection attempt (O'Brien & Associates)0.0s
PASSZero-quantity order rejected before DB0.0s
PASSHealthClaw patient -> encounter -> vitals -> diagnosis1.2s
PASSCross-platform: Telegram record visible in Webclaw0.8s
PASSPayroll run: 5 slips, $63,500 total, GL posted0.6s
PASSIndia GST inter-state 18% on Rs 50,000 = Rs 9,0000.1s
PASSCanada CPP below $3,500 exemption = $0 contribution0.1s
PASSUK PAYE on 50K: 14.97% effective rate0.2s
PASSEU reverse charge DE->FR: buyer self-assesses 20%0.1s
PASSBot self-recovered from missing naming series0.3s
PASSBot proactively flagged GL mapping error0.5s
PASSFINAL: $1,060,433.00 balanced, all gates passed0.4s
$

Tests Passing

Pass Rate %

Gates Passed

GL Breaks

## Every dollar balanced
$1,060,433.00 flowed through the general ledger during testing. Sales, purchases, payroll, expenses, asset depreciation, cancellations, reversals. The trial balance never broke, not once. All gates passed.
$0

$0

$250K seed

$500K sales

$850K payroll

$1.06M final

$1,060,433.00 DR = $1,060,433.00 CR, balanced to the cent

## We tried to break it
SQL injection, zero quantities, negative amounts, float precision, unicode edge cases, double-submits. Every attack vector we could think of. The system held.

$9,999,999.99 invoice
Decimal precision preserved

Quantity: 0
Rejected before DB

Quantity: -5
Rejected before DB

Price: 10.999 (float)
Forced to Decimal, rounded

SQL: O'Brien & Co
Parameterized, no injection

Unicode: Muller GmbH
Stored and rendered correctly

Delete submitted invoice
Blocked, reversal only

Double-submit draft
Blocked, must be draft

Won deal stage change
Terminal state enforced

Exceed leave balance
Correctly rejected

500+ operations
Trial balance still balanced

Cancel $9.9M invoice
GL reversals, books balanced

## End-to-end workflows
Not isolated unit tests, complete business workflows tested from start to finish. Every entity created, every GL entry posted, every cross-module handoff verified.

Order to CashClinicalAcademicProperty
Step 1
Customer
Wayne Enterprises

Step 2
Sales Order
2 line items, $140K

Step 3
Submit
Stock validated first

Step 4
Invoice
INV-2026-00007

Step 5
GL Entries
6 entries posted

Step 6
Stock Ledger
2 movements

Step 7
Payment
$140K received

Step 8
Trial Balance
Still balanced

All steps verified, Telegram + Webclaw cross-platform

Step 1
Patient
John Doe

Step 2
Appointment
Dr. Strange

Step 3
Encounter
Outpatient

Step 4
Vitals
7 measurements

Step 5
Diagnosis
J20.9 Bronchitis

Step 6
Prescription
Azithromycin 500mg

Step 7
Lab Order
CBC, STAT

Step 8
Insurance Claim
CPT 99213, $150

All steps verified, Telegram + Webclaw cross-platform

Step 1
Student
Alice Johnson

Step 2
Academic Year
2025-26

Step 3
Program
General Studies

Step 4
Course
MATH101

Step 5
Section
001, Prof Banner

Step 6
Enrollment
Enrolled

Step 7
Attendance
Present

Step 8
Grade
87/100 = B

All steps verified, Telegram + Webclaw cross-platform

Step 1
Property
Stark Tower

Step 2
Unit
Suite 100

Step 3
Lease
$5K/month

Step 4
Rent Schedule
12 months

Step 5
Work Order
HVAC, Urgent

Step 6
Rent Payment
$5K check

Step 7
GL Posted
Dr. Cash / Cr. AR

Step 8
Tenant
Peter Parker

All steps verified, Telegram + Webclaw cross-platform

## Two interfaces, one truth
51 cross-platform tests. Every record created via Telegram chat appeared correctly in the Webclaw dashboard. Instantly. Across all 4 verticals.

Telegram Chat
Create an invoice for Wayne Enterprises, $140,000
Done. INV-2026-00007 created. 6 GL entries posted. Stock updated.

Webclaw Dashboard
Invoice
Customer
Amount
Status
INV-00007
Wayne Ent.
$140,000
Submitted

Appeared instantly

51 cross-platform tests|ERPClaw, HealthClaw, EduClaw, PropClaw|100% pass rate

## Regional compliance tested
Tax calculations, payroll deductions, document validation, tested against real regulatory rules for 4 regions.

CA
Canada
HST, QST, CPP, EI, Federal Tax, BN Validation
12 tests passed

IN
India
GST Intra/Inter-state, GSTIN, TDS, PF
8 tests passed

GB
United Kingdom
VAT 20%, PAYE, NI
4 tests passed

EU
European Union
VAT, Reverse Charge, IBAN Validation
6 tests passed

## AI that catches mistakes
The AI doesn't just execute commands. It validates, self-corrects, and proactively flags errors, even when you don't ask.

Self-Recovery
Bot hit a missing naming series error. Instead of failing, it created the missing module, registered the series, and completed the operation. 5 self-recovery instances during testing.

Prerequisite Chains
Asked to enroll a student. Bot automatically created: Academic Year, Term, Program, Course, Section, Room, and Instructor, 8 objects in a single conversational flow.

Proactive Error Detection
While generating a P&L report, the bot flagged that payroll $63,500 was posted to Bank Charges instead of Salaries & Wages. Nobody asked, it caught the GL mapping error on its own.

## See the code. Run the tests.
ERPClaw is open source under the Free & Open Source license. Every test, every action, every GL posting rule is available for audit.
Related: read the architecture in ERPClaw OS, the trust posture in security, the open data and reproducible studies in research, or the install steps in core docs.
Get Started Free View on GitHub

---

## AI ERP Research: Open Data, Reproducible Studies
URL: https://www.erpclaw.ai/research/

# ERPClaw Research
Original research on the ERP, AI, and accounting markets. Open data, reproducible methodology, free to cite.

## Current research
Annual, Updated May 2026

### AI ERP Transparency Index 2026
A public scorecard of 50 ERPs measured against 12 transparency criteria for AI-native architecture. Open data, sourced scoring, reproducible methodology. Methodology locked May 2026; full ranking publishes Q3 2026.
Read the methodology &rarr;
Methodology

### The 5-Trait Test for AI-Native Architecture
The five architectural traits that separate AI-native ERPs from AI-decorated overlays. Free methodology, reproducible scoring, open challenge process. Used as the foundation of the Transparency Index.
Read the methodology &rarr;

## How to cite this research
All ERPClaw research is free to cite for any purpose, including commercial, academic, and journalistic use. We ask only that you cite us correctly so readers can find the source.

### Standard citation format
AvanSaber Inc. 2026. [Title of Research]. Available at https://www.erpclaw.ai/research/[slug]/

### Example
AvanSaber Inc. 2026. The 2026 AI ERP Transparency Index. Available at https://www.erpclaw.ai/research/ai-erp-transparency-index/

For academic citation in BibTeX, IEEE, or APA format, email support@avansaber.com and we will send the formatted entry.

## Why we publish research
Gartner's Magic Quadrant ranks ERP on overall capability. IDC reports sit behind a $5,000 paywall. Industry analysts disclose vendor relationships in footnotes that nobody reads. We publish research the opposite way: open data, open methodology, free downloads, and a documented challenge process for any vendor who disputes a score.
ERPClaw is a vendor in the market we measure. We disclose that on every page, we self-score using the same rubric we apply to competitors, and we publish our own scores even when they are not flattering. If you find a methodology error or a sourcing gap, file an issue on the public GitHub repo for the artifact in question. We correct in public.
Background on the architectural argument that drives this work: read the AI-native ERP pillar.

## Stay current on new research
Two artifacts published in 2026 so far. Three more queued for the next four quarters. Subscribe for release notifications.
Subscribe for release alertsRead the Transparency Index

---

## The 2026 AI ERP Transparency Index
URL: https://www.erpclaw.ai/research/ai-erp-transparency-index/

# The 2026 AI ERP Transparency Index
Scoring 50 ERPs Against the 5-Trait Test for AI-Native Architecture
By AvanSaber Inc. · Methodology locked 2026-05-24 · Full ranking Q3 2026

!
Status: In progress
Methodology locked May 2026. Per-vendor scoring in progress. The full ranking publishes when all 50 vendors are scored against all 12 criteria, every score sourced, with the public GitHub repo live alongside this page.

## What this is
A public, reproducible scorecard of the 50 ERPs that claim AI capabilities in 2026, measured against 12 transparency criteria. Each ERP gets a score from 0 to 100. Every score is backed by a public source. Methodology and data are open so any analyst, journalist, or buyer can verify a score or rerun the entire scoring against new evidence. The Index publishes annually; the next edition lands May 2027.
"Gartner ranks ERPs on capability and IDC puts its reports behind a $5,000 paywall. Neither one scores whether a vendor's AI claim is architecturally true. That gap is the whole reason this Index is free, open, and reproducible: a buyer should be able to check the receipt, not trust the ad."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## The 12 criteria
Each ERP is scored 0 to 10 on each of these 12 dimensions, then the 12 raw scores are summed and normalized to a 0 to 100 score. Criteria extend the established 5-trait test for AI-native architecture (see the methodology page) by adding pricing, sovereignty, model disclosure, and architectural recency.

### AI writes to the GL directly
Can the AI agent post journal entries autonomously, or does every action need a human to approve and submit?

### Pre-AI data model
Was the database schema designed before AI existed, or was it built with AI agents in mind from day one?

### Native action layer
Is there a programmatic action surface the AI calls directly, or does the AI translate intent into form-fills?

### Pre-write invariant enforcement
Are GL safety checks (debits equal credits, period open, accounts exist) enforced before the write, not patched up after?

### Single AI tier (no gating)
Is the AI feature set available to every customer, or gated behind the enterprise pricing tier?

### Open documentation of AI behavior
Is the prompt, behavior model, and decision logic documented publicly, or treated as a black box?

### Reproducibility of AI outputs
Can a customer replay an AI action and inspect the exact inputs, outputs, and GL entries produced?

### Foundation model transparency
Does the vendor name the LLM that powers their AI, and pin a version, or keep the model selection opaque?

### Customer-owned data and model isolation
Does customer data ever leave the vendor tenant for AI training? Opt-in, opt-out, never, or shared by default?

### Pricing transparency for AI
Is AI feature pricing published on the website, or only available through a custom sales quote?

### Open-source posture
Is the AI layer open source so customers can audit it, or is it closed code only the vendor can read?

### Year of last meaningful AI architecture change
When did the vendor last revise their AI architecture in a substantive way (not a marketing rebrand)?

## How scoring works

- ·0 to 10 per criterion. Each of the 12 criteria gets an integer score from 0 (the dimension is absent or undisclosed) to 10 (the dimension is fully present and publicly documented).
- ·Sources required. Every score must be backed by a public source: vendor documentation, press release, RFP response, marketplace listing, or product walkthrough. The source URL is recorded in the dataset.
- ·"Unknown" is treated as 0. If a dimension is undisclosed by the vendor and we cannot find a public source, the score is 0 with a footnote. Vendors can challenge this by submitting a public source.
- ·Normalization to 0 to 100. Raw scores (out of 120) are normalized to a 0 to 100 final score for readability.
- ·Vendors can challenge. Any score can be disputed by filing a public issue on the GitHub repo with a new public source. See the FAQ for the challenge process.
- ·Snapshot dated. Each annual edition locks a snapshot date. Architectural changes after the snapshot roll into the next edition, not the current one.

## The 5 tiers
The 50 vendors split across 5 tiers by segment and license model. Tier composition is published; the per-vendor roster within each tier is locked at methodology lock and published with the full ranking in Q3 2026.
| Tier | Segment | Count | Description
| Tier 1 | Enterprise | 10 | Large multi-entity ERPs serving Fortune 1000 buyers.
| Tier 2 | Mid-market | 15 | Mid-market ERPs and finance specialists serving 100 to 5,000 person companies.
| Tier 3 | SMB | 10 | SMB bookkeeping and small-business accounting platforms.
| Tier 4 | AI-native startups | 10 | The 2024 to 2026 cohort of finance and ERP startups built AI-first.
| Tier 5 | Open source | 5 | Open-source ERP and accounting projects with public source repositories.

Total: 50 vendors. Tier roster published with the full ranking in Q3 2026.

## Reproducibility
The Index is built to be reproducible by any third party. When the full ranking publishes in Q3 2026, the following will be public:

- ·All scoring data on GitHub at `avansaber/ai-erp-transparency-index` in JSON and CSV.
- ·Per-cell sources cited with public URLs.
- ·Methodology document in Markdown, version-controlled in the same repo.
- ·Reproducibility script that re-runs the scoring against any vendor's published documentation, so a third party can verify or contest a score.
- ·Public challenge process: file an issue with a counter-source, get a public response within 14 days, see corrections in the changelog.
Licensed CC BY 4.0. Free to cite, free to fork, free to build on. Attribution required.

## Annual cadence
The Index publishes once per year. The 2026 edition methodology locked May 2026; the full ranking publishes Q3 2026. The 2027 edition will publish May 2027 on the same scoring rubric, with the snapshot date and any methodology revisions documented in the changelog.
Between editions, vendor architectural changes get noted but do not retroactively change the current edition's scores. Major mid-year revisions (a vendor ships an architecture change that meaningfully alters a score) may be published as a notable update outside the annual cadence; these are clearly marked and dated.

## How to cite
Citation
AvanSaber Inc. 2026. The 2026 AI ERP Transparency Index. Available at https://www.erpclaw.ai/research/ai-erp-transparency-index/

For BibTeX, IEEE, or APA-formatted citation entries, email support@avansaber.com.

## Frequently asked questions

### Why these 12 criteria specifically?
Each criterion has a single answer the vendor can either prove with public evidence or cannot. We avoided subjective dimensions (ease of use, support quality, brand sentiment) because those are not auditable. The 12 dimensions extend the established 5-trait test for AI-native architecture by adding pricing transparency, customer data sovereignty, foundation-model disclosure, and architectural recency. Each adds a separable axis that vendors either publish or do not.

### Why does ERPClaw publish this?
Two reasons. First, the category is genuinely under-served. Gartner Magic Quadrant ranks ERP on overall capability, not on AI-architecture honesty. IDC reports cost $5,000 and sit behind a paywall. A free, reproducible, open scorecard is a hole in the market. Second, ERPClaw is a vendor in this category. We disclose that openly, we self-score using the same rubric, and we publish our own score even when it is unflattering. Independent journalists and academics can rerun the methodology and verify or dispute any score.

### How can a vendor challenge a score?
File a public issue on the GitHub repo for the artifact (avansaber/ai-erp-transparency-index, published at launch). Include the criterion number, the proposed new score, and a public source (vendor documentation, press release, RFP response, or product walkthrough) that supports the change. We respond within 14 days. Disputed scores get a public correction and a footnote in the next annual edition. Every revision is documented in the changelog.

### Will ERPClaw self-score?
Yes. Self-scoring is required for the methodology to be credible. Where ERPClaw scores high (open source, single AI tier, pre-write invariant enforcement) we publish the score. Where ERPClaw scores low (criterion 12 if applicable, since we are a younger product than NetSuite or SAP S/4HANA) we also publish the score and a note about what we plan to do. Honest self-scoring is the trust signal that makes the whole artifact credible.

### What if a vendor changes architecture mid-year?
We lock a snapshot date for each annual edition. The 2026 edition will note its scoring snapshot date (likely Q2 2026). Architectural changes published after the snapshot are noted in the next annual edition, not retroactively backfilled into the current one. This keeps the dataset reproducible. Vendors who ship major changes can request an interim re-score, which we may publish as a notable update outside the annual cadence.

### Why is the data not on this page yet?
Methodology was locked in May 2026. Per-vendor scoring is in progress and takes 40 to 60 hours across 50 vendors. We could publish partial data and update incrementally, but a half-scored ranking is worse than a fully-scored one because readers and journalists treat it as final. The full ranking publishes when all 50 vendors are scored against all 12 criteria, all sources cited, with the public GitHub repo live alongside the page.

### How does this relate to the 5-trait test?
The 5-trait test is the architectural foundation. The 12 transparency criteria extend it by adding dimensions that matter to a buyer beyond pure architecture: pricing transparency, model disclosure, data sovereignty, and architectural recency. Read the methodology page at /research/ai-native-vs-decorated-five-trait-test/ for how the 5 traits map to the 12 criteria.

---

## The 5-Trait Test for AI-Native ERP
URL: https://www.erpclaw.ai/research/ai-native-vs-decorated-five-trait-test/

# The 5-Trait Test for AI-Native ERP
Methodology for distinguishing AI-native architecture from AI-decorated overlays
By AvanSaber Inc. · Methodology version 1.0 · 2026-05-24

## Why this test exists
Every ERP vendor in 2026 claims "AI-powered." SAP markets Joule. Oracle markets AI Agents. Microsoft markets Copilot inside Dynamics 365. Sage, NetSuite, QuickBooks, Xero, Zoho, ERPNext, Odoo, every vendor in the category has an AI page. Most of those claims are AI-decorated overlays on forms-and-workflows systems designed before AI existed. A few are genuinely AI-native by architecture.
Buyers evaluating the category in 2026 cannot tell the two apart from marketing copy. The marketing copy is identical: every vendor says "AI-powered ERP." The architectural reality is not identical. This methodology is the test that tells them apart, expressed as five architectural traits a buyer can verify with public evidence.
This methodology is the foundation of the 2026 AI ERP Transparency Index, which extends the five traits into 12 measurable criteria scored across 50 vendors.
"Every ERP in 2026 has an AI page. Almost none of them changed the architecture underneath it. The five-trait test is the one question that separates the two: can the AI post to your general ledger by itself, inside a validated transaction, or does a human still click Submit on a form?"
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

## The five traits
Each trait is binary at the architectural level: the system either has it or does not. Vendors land in between because they have partial implementations or in-progress changes. The scoring rubric below covers the 10 (full), 5 (partial), and 0 (absent) scoring anchors for each trait.

### AI writes to the GL directly

An ERP that needs human approval on every AI suggestion is a smarter search box, not an AI agent. The architectural test is whether the AI can autonomously submit a state-mutating action with an audit trail, or whether every write still ends at a human clicking Submit on a form.

### Schema designed for AI agents

A schema designed in the 1990s assumes a human is driving every write. A schema designed for AI agents has the action layer as the API, immutable audit logs as a primary table type, and the foreign-key topology built so an agent can traverse it without help.

### Native action layer

If the AI translates intent into form-fills, it is a fancy text-to-form translator. If there is a programmatic action surface (an action per business operation, callable from chat, CLI, or web) the AI invokes the action directly. The latter is the architectural commitment.

### Pre-write invariant enforcement

An AI agent that posts an unbalanced journal entry can corrupt the books in seconds. The architectural defense is to enforce GL invariants (debits equal credits, period is open, accounts exist, currency matches) before the write, in the same transaction, with a clean rollback on failure.

### Single AI tier (no gating)

If the AI is the architecture, every customer gets it. If the AI is an upsell, the vendor is positioning it as a premium feature on top of a non-AI base product. The pricing structure reveals which one is true.

## Scoring rubric per trait
Each trait scored 0, 5, or 10. Intermediate scores allowed at the analyst's discretion when evidence supports a half-step (e.g., 3 if mostly absent with a single partial implementation, 7 if mostly present with one major gap). Every score must cite a public source.

### Trait 1: AI writes to the GL directly
10 / 10 (fully AI-native)
The AI invokes a state-mutating action through a programmatic surface. The action runs in a single transaction with full validation, posts to the GL, and writes an immutable audit row recording the AI's invocation, inputs, outputs, and exact GL entries. The customer can replay the action later.

5 / 10 (partial)
The AI prepares a draft journal entry or a pre-filled form. A human reviews and clicks Submit. The audit log records the human submit, not the AI's role in preparing it. Practical for early adopters who do not yet trust autonomous AI; not autonomous architecture.

0 / 10 (AI-decorated or absent)
The AI summarizes or answers questions about data. It does not write anything back to the books. Read-only assistant; no write path at all.

### Trait 2: Schema designed for AI agents
10 / 10 (fully AI-native)
Schema designed with AI agents as a primary user. Action layer is the API. Money stored as Decimal in TEXT (not float). IDs are UUIDs. GL is immutable by schema (no updated_at). Audit log is a first-class table that the AI writes to on every invocation. Foreign keys enforced. The AI can introspect tables to answer questions.

5 / 10 (partial)
Pre-AI schema with AI-related tables added in a recent release. The original tables (customers, invoices, GL entries) still assume human-driven writes. The AI-related tables (conversations, intents, suggestions) are bolted alongside. Backward-compatible but not native.

0 / 10 (AI-decorated or absent)
Pre-AI schema only. No tables related to AI invocations, AI audit, or agentic workflows. The vendor's AI lives entirely outside the database (a chatbot in the UI), not as a participant in the data model.

### Trait 3: Native action layer
10 / 10 (fully AI-native)
Every business operation is an action: kebab-case named, single-transaction, JSON in JSON out, fully documented in a module manifest. The AI calls the action by name. The same action runs from chat, CLI, web, or a programmatic API. There is no form-fill translation path.

5 / 10 (partial)
Some operations have actions; many still go through forms. The AI partially invokes actions for read operations and form-fills for write operations. Mixed architecture.

0 / 10 (AI-decorated or absent)
No action layer. Every operation is a form. The AI's only write path is to pre-fill forms for a human to submit. Adding AI does not change the architecture; it just adds a smarter input method.

### Trait 4: Pre-write invariant enforcement
10 / 10 (fully AI-native)
Every GL posting runs through a multi-step validation pipeline before any row is inserted. Debits equal credits, period open, accounts exist, currency match, etc. Failure aborts the entire transaction. The AI cannot violate the GL even when it is wrong about something else. Invariants are documented and testable.

5 / 10 (partial)
Some invariants enforced at write time. Some enforced by a post-write batch reconciliation job. Imbalanced entries can land temporarily and get caught later. Practical for legacy systems; not safe for autonomous AI.

0 / 10 (AI-decorated or absent)
No invariant enforcement at the data layer. Every safety check lives in the UI or in a nightly batch. An AI agent can post any row that fits the schema. Books can corrupt; cleanup is manual.

### Trait 5: Single AI tier (no gating)
10 / 10 (fully AI-native)
AI is available to every customer at every tier, free or paid, no upcharge. The vendor's pricing page does not separate AI features as a separate line item. The AI is the product, not an add-on.

5 / 10 (partial)
AI available in mid-tier and above, gated out of the entry-level plan. Vendor positions AI as a paid premium feature.

0 / 10 (AI-decorated or absent)
AI only available in the enterprise tier, custom-quoted, gated behind a sales call. AI is a $50K+ annual add-on positioned as the premium offering on top of a forms-based base ERP.

## How to apply the test to any ERP
Six-step process. Roughly 60 to 90 minutes per vendor for the first pass; faster once you have the rhythm.

- 1.Read the vendor's AI page. Capture the claims verbatim. The vendor's marketing language is the input, not the conclusion.
- 2.Find their developer documentation. Look for an action / API / SDK page. The presence and shape of the programmatic surface is the strongest signal for trait 3 (native action layer).
- 3.Watch a product demo or walkthrough. Pay attention to whether the AI invokes an action or pre-fills a form. The architectural answer is in the workflow, not the marketing.
- 4.Check the pricing page. Single AI tier vs gated AI tier vs enterprise-only AI = trait 5.
- 5.Score each trait 0, 5, or 10. Cite a public source per score. Sum and normalize: total / 50 * 100 = trait score (0 to 100).
- 6.Publish the score and the sources. If you are scoring publicly, accept the open challenge process: vendors may submit corrections backed by counter-sources.

## Common misapplications
Buyers who work through this test for the first time often hit five recurring patterns where a vendor's marketing makes a partial implementation look like a full one. These are not edge cases; they show up in the majority of first-pass evaluations.

### Treating "AI suggests, human approves" as passing Trait 1
Trait 1 tests autonomous write capability, not suggestion quality. A workflow where the AI prepares a journal entry and a human clicks Submit is precisely what Trait 1 is designed to disqualify. The suggestion can be sophisticated, context-aware, and correct nine times out of ten; it still scores 5, not 10. The architectural question is narrow: can the AI invoke a state-mutating GL action without a human in the critical path? If the answer is no, the score is 5 at most, regardless of how intelligent the suggestion appears to the user.

### Conflating a REST API with a native action layer (Trait 3)
Many ERPs expose a REST or GraphQL API for data sync and integration. A generic CRUD endpoint is not a native action layer. Trait 3 looks for a named action surface the AI calls by business operation: approve-invoice, recognize-revenue, post-accrual. The test is whether the AI calls an operation by name in a single-transaction, JSON-in JSON-out call, or whether it constructs raw HTTP requests against a CRUD endpoint to achieve the same result. The former is native. The latter is a workaround. Vendors with external integration APIs often claim Trait 3 on that basis; the workflow test in step 3 of the application guide will reveal which is which.

### Using a bolted-on AI schema to claim Trait 2 compliance
Vendors who add AI-adjacent tables (conversations, intents, suggestions) to a pre-2020 schema sometimes score themselves high on Trait 2. The test is not whether AI-related tables exist; it is whether the core accounting tables were designed with AI as a primary user. The tell is in the original GL schema: money stored as a float rather than a fixed-precision type, mutable GL entries with an updated_at column, foreign keys unenforced at the database level, and no immutable audit table for AI invocations. A conversations table added in 2024 sits on top of an architecture that was not designed for agents, and Trait 2 reflects the underlying design, not the layer added later.

### Scoring Trait 5 on plan names instead of pricing page evidence
"AI Professional" and "AI Business" are marketing labels, not pricing signals. Trait 5 has a single test: can a customer on the entry-level plan invoke an AI action that writes to the GL? The pricing page is the source of truth. If the AI write capability is absent from the lowest paid tier, Trait 5 scores 0 or 5 regardless of what the plans are named. Check the feature comparison table, not the plan name. The feature comparison table is where vendors hide the gate.

### Missing the transaction requirement in Trait 4
Pre-write invariant enforcement is only meaningful if the validation and the write execute in the same database transaction. Some vendors enforce GL balance rules in application code before issuing a write, but the validation and the commit are separate calls. A concurrent write can land an unbalanced entry in the window between them. The architectural test is whether a validation failure triggers a rollback before any row is inserted. Batch reconciliation that catches imbalances after the write scores Trait 4 at 5; it catches errors but does not prevent them. The distinction matters for autonomous AI writes, where a correction loop is more expensive than a pre-write gate.

## How the 5-trait test extends to the 12-criteria Index
The 5 traits cover architecture. The 12 criteria in the Transparency Index cover architecture plus four additional dimensions that matter to a buyer beyond pure architecture: pricing transparency, customer data sovereignty, foundation-model disclosure, and architectural recency. The mapping:
| 5-trait test | Index criteria covered
| Trait 1 (AI writes to GL) | Criterion 1
| Trait 2 (Schema for AI) | Criterion 2
| Trait 3 (Native action layer) | Criterion 3, plus reproducibility (criterion 7)
| Trait 4 (Pre-write invariants) | Criterion 4
| Trait 5 (Single AI tier) | Criterion 5, plus pricing transparency (criterion 10)
| Added in 12-criteria Index | Criteria 6 (open AI docs), 8 (foundation-model disclosure), 9 (data sovereignty), 11 (open source), 12 (architectural recency)

The Index uses the 12-criteria rubric because buyers ask about all 12 dimensions. The 5-trait test is the cleaner version when an analyst wants the architectural answer only.

## Open challenge process
Any vendor scored under this methodology may request a re-scoring with new evidence. The process is public and documented so the methodology stays credible.

- 1.File a public issue. On the GitHub repo for the artifact being challenged (for the Index, this will be `avansaber/ai-erp-transparency-index` when the Index publishes in Q3 2026).
- 2.Cite the trait or criterion number. Specify which score you are disputing and what you propose as the corrected score.
- 3.Provide a public counter-source. Vendor documentation, press release, RFP response, product walkthrough video, or live demo. Private NDA evidence is not acceptable because the methodology requires reproducibility.
- 4.We respond within 14 days. If the counter-source is strong, we update the score, publish a correction in the changelog, and note the revision in the next annual edition.
- 5.All revisions are public. The changelog records every score change, the reason, and the source that triggered the change. The dataset history is preserved in git.
ERPClaw is a vendor in this category. AvanSaber Inc. self-scores using this methodology and accepts public challenges on its own scores the same way as on competitor scores. Self-scoring transparency is the trust signal that makes the methodology credible.
"We run these five traits on ERPClaw itself, publish the score, and take public challenges on it, exactly as we do for NetSuite or SAP. A test you will not turn on your own product is marketing, not measurement."
Nikhil Jathar, co-founder, ERPClaw (AvanSaber Inc.)

---

## Security - Self-Hosted, Local-First, Zero Cloud Data
URL: https://www.erpclaw.ai/security/

# Your data, your rules
ERPClaw is designed from the ground up with a security-first architecture. Self-hosted, zero network access, and a complete immutable audit trail.

## Security Principles

### Self-Hosted by Default
Your ERP runs on your infrastructure. No data ever leaves your network. No cloud dependency, no vendor lock-in, no third-party access.

### Zero Network Access
ERPClaw skills make no outbound network calls. No telemetry, no analytics, no phone-home. Your data stays exactly where you put it.

### Immutable Audit Trail
General ledger and stock ledger entries are never modified or deleted. Cancellations create reversal entries, preserving a complete audit history.

### SQL Injection Safe
Every database query uses parameterized statements. No string interpolation, no dynamic SQL construction, no injection surface.

### No Credentials Required
ERPClaw uses only Python standard library modules (sqlite3, json, decimal, uuid). No API keys, no OAuth tokens, no secrets to manage.

### Atomic Transactions
Every submit operation (GL posting, stock movement, invoice) runs in a single SQLite or PostgreSQL transaction. Any failure triggers a full rollback.

## How It Works

### Single-File Database (default)
By default, all data lives in one SQLite file (~/.openclaw/erpclaw/data.sqlite). Backup is copying a file. PostgreSQL is fully supported as an alternative backend via PyPika; same code, same actions.

### WAL Mode Isolation
SQLite Write-Ahead Logging provides read/write concurrency without locking on the default backend; PostgreSQL provides MVCC concurrency. Foreign key constraints enforced at the database level on both.

### Decimal Precision
All financial amounts stored as TEXT and processed through Python's Decimal library with ROUND_HALF_UP. No floating-point rounding errors.

### Open Source
Full source code available for audit. No obfuscation, no proprietary components. Review every line of code that touches your financial data.

## ERPClaw OS: AI Safety Built In

### Constitutional Articles
Every operation is governed by inviolable articles. Financial integrity, double-entry enforcement, immutable ledgers, and data ownership rules that cannot be overridden, not even by the AI itself. Automated tests verify compliance.

### 11 Protected Financial Files
The DGM (Deterministic Generation Model) has a hard-coded exclusion list of 11 financial files that can NEVER be modified by the AI. gl_posting.py, stock_posting.py, tax_calculation.py, and other core financial logic are permanently protected.

### Invariant Checks Per Batch
Every transaction batch is validated against invariant checks: GL double-entry balance, voucher balance, immutability verification, and cross-table consistency. Failures block the entire batch.

### Tier Classification System
Modules are classified into 4 tiers (0-3) based on deployment autonomy. Tier 0 (financial core) requires manual approval for any change. Tier 3 (low-risk utilities) can be autonomously updated. The classification itself is immutable.

## ERPClaw vs. Cloud ERP Security
| Aspect | ERPClaw | Cloud ERPs
| Data location | Your server, your network | Vendor's cloud
| Network calls | Zero (no internet required) | Constant (SaaS model)
| Audit trail | Immutable, append-only | Vendor-managed logs
| Source code | Fully open, Free &amp; Open Source | Proprietary, closed
| Vendor access | None | Admin access to your data
| Encryption | Your choice (disk/network) | Vendor-managed keys
| Compliance | You control everything | Shared responsibility

## Audit it yourself
Every line of code is open source. Review our security model, run your own penetration tests, and deploy with confidence.
Related: see the test surface at quality, install steps at core docs, or pricing posture at pricing.
View Source Code

---

## ERPClaw support: GitHub, Discord, email, paid services
URL: https://www.erpclaw.ai/support/

# How to get help
Community-tier through GitHub and Discord. Direct email to the co-founder. Paid implementation and ongoing-support contracts available. Pick the channel that fits your question.

## Support channels

### GitHub Issues
Public bug reports, feature requests, and questions. Most product issues are resolved here within 1 to 5 business days. Search existing issues before opening a new one. ERPClaw maintainers triage daily.
Best for: bug reports with reproducible steps, feature requests, questions answerable in async public discussion.

### GitHub Discussions
Long-form architectural questions, integration patterns, customization help, and showcase. Community + maintainers participate. Slower than Discord but higher signal for design conversations.
Best for: 'how should I model this?' questions, integration design, sharing what you built.

### Discord
Real-time chat with maintainers and the community. The fastest path for a quick question or a stuck-on-install moment. Active during US Pacific business hours; async coverage from international community.
Best for: quick questions, install help, real-time troubleshooting.

### Email AvanSaber Inc.
Direct line to Nikhil (co-founder). Use this for confidential issues, business-relationship questions, or matters that don't fit a public channel. Response within 1-2 business days.
Best for: confidential questions, paid implementation interest, partnership inquiries.

### Paid implementation services
AvanSaber Inc. offers paid implementation, migration, and ongoing-support engagements for teams that prefer managed help. Pricing scales with scope; typical engagements run 4-12 weeks elapsed. Email Nikhil to scope.
Best for: mid-market teams with a defined timeline who want a vendor on the hook.

## Frequently asked questions

### What is the response time for an open-source ERP support question?
GitHub Issues: 1 to 5 business days for triage. Discord: same-business-day for active hours, longer for async. GitHub Discussions: variable, depending on question complexity. Email: 1 to 2 business days. There is no SLA on community-tier support; if your business requires guaranteed response times, email about a paid support contract.

### Is community support enough for a real production deployment?
For most SMB-to-mid-market deployments, yes. ERPClaw is on the Stripe Marketplace with paying customers running real production books on community-tier support. The architectural advantages (an immutable ledger, constitutional rules on every operation, layered automated testing) reduce the support surface significantly compared to a SaaS where opaque vendor cloud issues require vendor cooperation. For Fortune 500 deployments with strict SLA requirements, paid support contracts are available.

### Does ERPClaw offer 24/7 enterprise support?
Not as a default community-tier offering. Paid support contracts from AvanSaber Inc. can include 24/7 coverage; pricing scales with the SLA requested. For SMB-to-mid-market deployments, the typical setup is community-tier with optional paid support for specific incidents.

### What about legal liability or warranty?
ERPClaw is provided under Free &amp; Open Source without warranty (standard open-source disclaimer). Paid implementation contracts from AvanSaber Inc. can include defined warranty terms for the implementation work; the underlying open-source software remains as-is. Most teams handle this the same way they handle Linux on production servers: open-source software plus optional paid support.

### Where do I report a security vulnerability?
Email support@avansaber.com with the details. We follow responsible disclosure: report privately, give us a reasonable window to fix, then coordinate public disclosure. We do not currently run a paid bug-bounty program but we credit security researchers in release notes.

Related: read the install guide, the migration playbooks, the test surface, the security overview, or the pricing page.

---