Build or change a headless store with a coding agent
One company's docs show a full fit.
What counts as done
A test order placed end to end on the test store (cart, payment session, completion, order webhook received) after the agent's change, with the agent able to authenticate to the docs or dev tools.
Measured, for our own tests: authentication succeeds on the first try. No docs we read state these figures, so they don't count against a listing here.
How agents ask for it: set up a Medusa backend for my shop
, wire the storefront cart to checkout
, add a payment provider
, fix the webhook that marks orders paid
Does the job 1
The docs show it meets the test above, and nothing we read says otherwise.
| Listing | Why |
|---|---|
| Spree Commerce BB Vendo Connect Inc | The docs MCP at spreecommerce.org/docs/mcp needs no sign-in and npx spree add ai wires Claude Code, Cursor and VS Code; a free hosted sandbox and webhooks for order.placed cover the test order. source |
Does the job, with a limit 16
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Adobe Commerce (Magento) D Adobe Inc. | Magento Open Source is free to self-host, but Adobe has no MCP or docs MCP for Commerce, and a first-hand account shows a community MCP module's OAuth discovery returning Magento's HTML 404 in Cursor. source |
| BigCommerce API + MCP B Commerce.com US, Inc. | The docs MCP at docs.bigcommerce.com/_mcp/server needs no sign-in. Free sandboxes are for partners only, others get a 15-day trial, and checkout happens in a browser, so the end-to-end test order is manual. source |
| Commerce Layer API + MCP B Commerce Layer, Inc. | The Documentation MCP at docs.commercelayer.io/~gitbook/mcp needs no sign-in, and the free Developer plan has unlimited test orders, but breaking changes have shipped with same-day notice. source |
| commercetools BB commercetools GmbH | A 60-day trial without a card covers core APIs, but Managed MCP Servers sit outside the standard SLA and the agent products are not in the trial (per the last assessment). source |
| Ecwid by Lightspeed C Ecwid, Inc. | Only paid plans reach the API and there is no sandbox (a free test-store upgrade is by email to support); no docs MCP was found, and cart and checkout live in the browser JavaScript API. source |
| Elastic Path API + MCP C Elastic Path Software Inc. | The Developer MCP returns storefront docs and code with a tokens cap (default 5,000), but it is in beta and needs an invite-only access key, so a coding agent cannot sign in on first try. source |
| Medusa API + MCP C MedusaJS, Inc. | The docs MCP is OAuth from Claude Code, Codex and Cursor but needs a Cloud account; a first-hand Claude Code sign-in failed on an issuer mismatch (fixed 30 September), and another account looped on broken payment-collection webhooks. source |
| Saleor API + MCP BB Saleor Commerce sp. z o.o. | Free non-commercial Cloud sandboxes and a Documentation MCP exist, but the dummy payment plugins were removed, so a test order needs a payment app, and docs MCP sign-in wasn't read. source |
| Shopify API + MCP BB Shopify Inc. | Dev MCP runs locally with no authentication from Claude Code and Cursor, and development stores are free, but Dev MCP answers of about 20,000 tokens filled context and password-protected stores block cart tools. source |
| Shopware BB shopware AG | Community Edition runs free in Docker with shopware-cli, but no docs MCP was found in the docs index, and the MCP is experimental with open issues on OAuth discovery. source |
| Snipcart API + MCP D Snipcart inc. | Free test mode with ST_ keys and order webhooks cover a test order, but no docs MCP exists and the cart is a browser widget a coding agent can only wire into HTML. source |
| Square B Block, Inc. | Sandbox access works only through the local server with SANDBOX=true; the remote server reaches production only, and Square's checkout is hosted rather than a headless storefront backend. source |
| Swell C Swell Commerce Corp. | Test environments and partner development stores exist and the MCP covers Test; no docs MCP was found, so a coding agent signs in to a store to use read_guide. source |
| Vendure BB Elevantiq GmbH | Self-hosting is free, but accounts show Console-gated development tools in Cursor never starting sign-in and OAuth refusing clients without the RFC 8707 resource parameter (both closed in November). source |
| Wix Stores and eCommerce API B Wix.com Ltd. | The remote MCP signs in by OAuth from Claude Code and installs in Cursor with one click, and demo carts give a test mode, but the site must be published for checkout. source |
| WooCommerce API + MCP BB WooCommerce, Inc. | Free self-hosted test stores and a documented headless Store API with cart tokens, but no docs MCP was found in the docs index, and 11.3 removed the old MCP endpoint. source |
Doesn't do this job 3
Listings in the category that do something else, with the reason.
| Listing | Why |
|---|---|
| Agentzon D Merit Systems Inc (agentzon.co terms); Merit Systems, Inc. (merit.systems terms) | Not a platform to build a store on. source |
| doola MCP F doola | Not a commerce backend. source |
| Fabric E Commerce Fabric, Inc. | Fabric is a merchant order-management and catalogue backend sold by contract; there is no self-serve test store, and its docs MCP only searches documentation. source |
What stops agents finishing
- MCP OAuth that common clients cannot complete
- docs or dev MCP responses too large for context
- breaking changes in minor releases
Around it
Before: code.repo-open. After: code.deploy, commerce.store-sync.
Where this comes from
Weight 11 of 100 in Commerce & checkout: 6 of 49 accounts (Medusa, Vendure and Magento issues, Hacker News, Shopify forum)
The same as data: /api/v1/use-cases/commerce.build-storefront.json, or the MCP tool get_use_case with id commerce.build-storefront. Other jobs in this category: /api/v1/niches?category=commerce.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.