Stripe Agent Toolkit
Stripe's official SDK and MCP server for wiring AI agents into payments, billing, and customer operations.
Engineering teams already on Stripe who want to give LLM agents controlled access to payments, invoicing, and billing without hand-rolling tool schemas or an MCP server.
Teams not using Stripe, or anyone who wants a generic multi-processor payments abstraction; also a poor fit where unattended agents must not touch money movement.
Stripe Agent Toolkit is Stripe's official open-source SDK and remote MCP server for wiring AI agents into Stripe's payments, billing, and customer APIs. It ships as a monorepo (TypeScript and Python) with three main entry points: @stripe/ai-sdk for Vercel's AI SDK, @stripe/token-meter for framework-free integrations with the OpenAI, Anthropic, and Google Gemini SDKs, and a hosted remote MCP endpoint at https://mcp.stripe.com that agents connect to over OAuth. There are also first-class installers for Claude Code, Codex, Cursor, and Grok Build so an agent gets a curated set of Stripe tools with a single command. Exposed capabilities cover the surface most commerce agents need: creating and looking up customers, products and prices, generating payment links and invoices, refunding charges, reading balance and dispute data, and running billing operations against subscriptions. Because it is a thin, typed wrapper over the Stripe API, restricted API keys map directly onto the tools an agent can call, which lets you scope an autonomous workflow to, say, invoice creation without granting refund powers. Common workflows include letting a support agent issue a refund from a chat, building an e-commerce copilot that produces payment links on the fly, running an internal ops assistant that pulls MRR and dispute stats, and metering LLM usage so an agent can bill its own users through Stripe. The MCP server is the fastest on-ramp for Claude Desktop, Cursor, and other MCP-aware clients; the SDKs are the better fit when you are shipping a production agent and want deterministic tool schemas inside your own framework.
This is the reference implementation for wiring Stripe into an agent, and it shows. The combination of a hosted MCP endpoint, a Vercel AI SDK plugin, and a framework-free SDK covers virtually every way people are building agents today, and the restricted-key model is the right primitive for scoping what an LLM is allowed to do. I would still gate any refund or payout tool behind a human confirmation, but as a starting kit this is the one I would reach for.
— The AI Tool Bible editorial team
Pros
- ✅ Official Stripe project, MIT-licensed, maintained in the stripe/ GitHub org rather than a community fork.
- ✅ Covers three integration styles: hosted MCP server, Vercel AI SDK plugin, and framework-free SDK for OpenAI/Anthropic/Gemini.
- ✅ One-command installers for Claude Code, Codex, Cursor, and Grok Build make it trivial to add Stripe tools to popular agent clients.
- ✅ Uses Stripe restricted API keys, so tool permissions map cleanly onto real Stripe scopes rather than a bespoke auth layer.
- ✅ Includes a token-meter helper so agents can bill their own LLM usage back to end customers through Stripe billing.
- ✅ Both TypeScript and Python bindings, keeping parity for teams standardized on either stack.
Cons
- ⚠️ Only useful if you already use Stripe; there is no abstraction for Adyen, Braintree, or other processors.
- ⚠️ Tool coverage tracks the Stripe API surface but is not exhaustive — advanced Connect, Radar, or Terminal workflows may still need direct API calls.
- ⚠️ Giving an autonomous agent a live Stripe key carries real financial blast radius; safe use requires careful restricted-key scoping and human approval steps.
- ⚠️ The remote MCP server centralizes auth on Stripe's endpoint, which means outages or rate limits there hit every connected agent.
- ⚠️ Documentation is spread across GitHub, docs.stripe.com/agents, and per-package READMEs, which slows first-time setup.
Use cases
Explore related
Compare with similar tools
All in MCP Servers →Airtable MCP Server
Model Context Protocol server that gives LLMs full read/write access to your Airtable bases.
Apple Notes MCP
Let Claude read your local Apple Notes over the Model Context Protocol.
AWS MCP Servers
Official AWS Labs collection of Model Context Protocol servers for connecting AI coding assistants and agents to AWS services and documentation.
Azure MCP Server
Official Microsoft MCP server that exposes 40+ Azure services to AI agents under the developer's own Entra ID credentials.
Blender MCP
Give Claude hands inside Blender — an MCP server for natural-language 3D modeling.
Brave Search MCP
Give any MCP client web and local search powered by the Brave Search API.