ai

This MCP server provides Stripe's Model Context Protocol access for integration and testing of MCP flows locally. You can run a local Stripe MCP server using npx, which exposes the MCP interface compatible with client libraries and agent tooling. The server is designed to work with Stripe's ecosystem, including the Stripe Agent Toolkit and related MCP-enabled tooling, enabling secure OAuth-based access and function-calling style interactions with Stripe APIs via MCP clients. Use the provided command to start a local server and connect your MCP clients with the appropriate Stripe secret key to authorize operations.

Once running, you can interact with the MCP endpoints using MCP clients or agent tooling that supports MCP. The documentation and examples in the repository cover how to fetch available methods, request permission scopes, and perform structured calls through the MCP format. The Stripe Agent Toolkit and related libraries (e.g., ai-sdk, token-meter) can surface tools from the MCP server for use in agent environments, enabling secure, scoped access to Stripe resources through your agent workflows.

Prerequisites:

• Node.js v18+ and npm installed
• An active Stripe account with API keys

Install and run locally:

1. Ensure Node.js is installed. Verify with: node -v npm -v

2. Start the MCP server locally using npx (no need to install the package globally): npx -y @stripe/mcp --api-key=YOUR_STRIPE_SECRET_KEY

3. Replace YOUR_STRIPE_SECRET_KEY with a Stripe secret key that has the desired permissions (preferably a Restricted API Key for security).

4. Once the server starts, use your MCP client tools to connect to the local server endpoint as documented in the MCP docs and the repository examples.

Tips and notes:

• Use a Restricted API Key (rk_*) for better security and granular permissions when interacting with the MCP server.
• The command expects you to provide a Stripe secret key via the --api-key flag; avoid hardcoding keys in code and consider environment-based loading in your deployment.
• If you need to run in a container or CI, consider setting the key via environment variables and constructing the command accordingly.
• Refer to the MCP documentation linked in the README for details on supported methods, OAuth flow, and tool integration with agent frameworks.
• The server name in this configuration is ai (Stripe AI MCP server). If you mirror multiple servers, expand the mcpServers map accordingly.