FastAPI-MCP turns an existing FastAPI application into an MCP server, auto-generating MCP tools from its endpoints while preserving request/response schemas and documentation. It calls itself "FastAPI-native" rather than a generic OpenAPI-to-MCP converter: it mounts directly onto your FastAPI app and communicates over FastAPI's ASGI interface instead of making HTTP calls back to your own API.

Use it when you already have a FastAPI application and want to expose its endpoints to an MCP client with minimal extra configuration - the basic setup is adding two lines to your FastAPI app. It's a fit if you want authentication for the MCP endpoints to reuse your existing FastAPI Depends() dependencies rather than a separate auth layer. It's not for exposing arbitrary REST APIs written in other frameworks - it works specifically with FastAPI apps via their native ASGI interface, though the resulting MCP server can still be deployed separately from the original app if you don't want them running in the same process.

• Authentication built in, reusing your existing FastAPI dependencies.
• Zero/minimal configuration - point it at a FastAPI app and it works.
• Preserves request/response model schemas and endpoint documentation exactly as they appear in your FastAPI/Swagger setup.
• Flexible deployment: mount the MCP server on the same app, or deploy it separately.
• Uses FastAPI's ASGI transport directly for efficient communication, without HTTP calls back to your own API.

Install via uv add fastapi-mcp (or pip install fastapi-mcp), then mount it directly on your FastAPI app:

from fastapi import FastAPI
from fastapi_mcp import FastApiMCP

app = FastAPI()

mcp = FastApiMCP(app)

# Mount the MCP server directly to your FastAPI app
mcp.mount()

This exposes the auto-generated MCP server at https://app.base.url/mcp. Requires Python 3.10+ (3.12 recommended) and uv. It is MIT licensed (Copyright 2025 Tadata Inc.), and a hosted, managed version is available at tadata.com for teams that don't want to self-host.

Teams that already run a FastAPI service and want to expose it to AI agents as MCP tools without maintaining a separate conversion layer or duplicating authentication logic. It requires Python 3.10 or newer (3.12 recommended) and the uv package manager, and a community Slack (MCParty) is available for questions and support. The project publishes full documentation and an examples directory covering these features in more depth for teams that need advanced setups.