Stripe runs an official MCP server at mcp.stripe.com that exposes the Stripe API as tools for AI models. Read customers and subscriptions, look up invoices and disputes, summarise revenue and trace payment failures — all from chat, with a restricted Stripe API key.
https://mcp.stripe.com/
Claude Sonnet 4.5
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.
Stripe restricted API key. Create one at dashboard.stripe.com/apikeys with the minimum read scopes you need; never paste a live secret key.
How models use it and what it is built for.
The Stripe MCP server wraps the Stripe API and surfaces it as MCP tools — listing customers and subscriptions, fetching invoices and charges, retrieving payment intents and disputes, and (with the right scopes) creating refunds or sending invoices. It is intended for back-office and analytics workflows: revenue rollups, churn investigation, billing support triage, dunning analysis. Combine it with a database MCP (Postgres, Supabase) for full customer-360 questions in chat.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
How many active subscribers did we have at the start of this month vs today?
List the 10 largest disputes this quarter and summarise the reasons.
Find this customer's payment history and tell me why their last invoice failed.
Refund invoice in_1ABC for the full amount and add a note that it was a duplicate charge.
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 Stripe
Claude Sonnet 4.5
Claude Sonnet 4.5 is reliable on multi-call billing investigations. Use Haiku 4.5 for high-volume read queries and GPT-5.2 for natural-language financial summaries.
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.
Try Stripe in Agent StudioCommon questions about connecting, scoping and using it safely.
It is an official MCP server hosted by Stripe at mcp.stripe.com. It exposes the Stripe API as tools an AI model can call — reading customers, subscriptions, invoices, payments and disputes from chat.
Use a restricted API key with read-only scopes for the resources the agent needs (customers, subscriptions, invoices). Avoid live secret keys. Stripe enforces scopes server-side, so a read-only key cannot create refunds even if the model tries.
Only if your restricted key has write scopes for that resource. By default we recommend starting with read-only access for analytics and support workflows, then adding narrow write scopes (e.g. refunds) once you trust the agent on a sample.
Yes. Any MCP-compatible client can connect to mcp.stripe.com — Claude Desktop, Cursor, Cline, browser tools like MCP Agent Studio. Paste the URL and your restricted key into the client to start.
Claude Sonnet 4.5 handles multi-call investigations (e.g. "trace this customer's payment failure") well. For high-volume read queries Haiku 4.5 is cheaper. GPT-5.2 is strong at financial-summary writing.