Square’s MCP server reaches the whole Square API through three tools rather than hundreds. An agent discovers the method it needs, checks its parameters, then makes the call — across payments, orders, customers, catalogue and the rest.
https://mcp.squareup.com/mcp
Claude Sonnet 5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
The hosted server uses OAuth with your Square account and acts on production. The local server takes an access token and can target the sandbox.
How models use it and what it is built for.
Instead of a tool per endpoint, the server has three: get_service_info lists the methods a Square service offers, get_type_info returns the parameters a method needs, and make_api_request executes the call. The model works the way a developer reading docs would: discover, check, then call. That keeps the tool list tiny while covering the full API, but it moves the hard part into the model, which has to build a correct request body from the type information. The hosted server uses OAuth and acts on production only. The local server, run with npx, can point at a sandbox and can be made read-only with DISALLOW_WRITES — the right setup while you see what an agent does with a payments API.
Typical tools an AI model can call. Exact names vary by version.
Taken from the official Square documentation — see Square MCP documentation for the full reference.
Environment variables
ACCESS_TOKENrequiredSquare access token (local server).
SANDBOXSet to true to use the sandbox environment.
PRODUCTIONSet to true to use production.
DISALLOW_WRITESSet to true to restrict the server to read-only requests.
SQUARE_VERSIONPin the Square API version, e.g. 2025-04-16.
Client configuration
Local, sandbox, read-only
The safest way to start. Square recommends testing prompts in a sandbox before production.
{
"mcpServers": {
"square": {
"command": "npx",
"args": ["square-mcp-server", "start"],
"env": {
"ACCESS_TOKEN": "<SQUARE_SANDBOX_TOKEN>",
"SANDBOX": "true",
"DISALLOW_WRITES": "true"
}
}
}
}Copy any of these into MCP Agent Studio after connecting.
What were total sales at the downtown location yesterday, by payment method?
Find the customer with email jane@example.com and list her last five orders.
Which catalogue items are out of stock at any location?
Show refunds over $100 issued this week and the orders they belong to.
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Square
Claude Sonnet 5
With only three generic tools, the model must build each request body from type information. Sonnet 5 follows the discover-then-call pattern and gets nested fields right.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
It reaches the whole Square API through get_service_info, get_type_info and make_api_request. The model discovers the method, checks its parameters, then calls it, instead of choosing among hundreds of tools.
Yes, with the local server: set SANDBOX=true and use a sandbox access token. The hosted server works on production only.
Run the local server with DISALLOW_WRITES=true. It then refuses write, update and delete requests.
Square labels it beta and keeps an allowlist of MCP clients for the hosted server. Test prompts in a sandbox before you point an agent at live data.
Stripe exposes a tool per operation. Square exposes three generic tools over its whole API, which covers more but asks more of the model. Square is the natural fit for in-person and point-of-sale businesses.
Stripe
Query customers, subscriptions and payments with an AI agent.
PayPal
Create invoices, capture orders, issue refunds and track disputes from an MCP client.
Cashfree
Query payments, orders, settlements and refunds on India’s payment rails.
HubSpot
Query CRM, deals and tickets — and draft follow-ups with full record context.