# FastMCP

> The fast, Pythonic framework for building MCP servers and clients that connect LLMs to tools and data.

FastMCP is the standard framework for building Model Context Protocol (MCP) applications in Python. It provides a clean, Pythonic way to create MCP servers and clients that connect large language models to tools and data. Originally incorporated into the official MCP SDK in 2024, FastMCP is now downloaded over a million times daily and powers approximately 70% of MCP servers across all languages.

The framework simplifies the complex process of building effective MCP servers by handling serialization, validation, error handling, and protocol compliance automatically. This allows developers to focus on business logic rather than protocol implementation details.

- **Components** expose tools, resources, and prompts to LLMs. Simply wrap a Python function with a decorator, and FastMCP automatically handles schema generation, validation, and documentation.

- **Providers** define where components originate from, supporting decorated functions, files on disk, OpenAPI specs, remote servers, and custom sources for maximum flexibility.

- **Transforms** shape what clients see through namespacing, filtering, authorization, and versioning. The same server can present differently to different users based on context.

- **MCP Proxy** enables connecting to and proxying remote MCP servers, allowing composition of multiple MCP services.

- **Authentication and Authorization** features provide built-in support for securing MCP servers with various authentication methods and fine-grained access control.

- **Multiple Transport Options** support stdio, HTTP/SSE, and WebSocket connections for flexible deployment scenarios.

- **LLM-Friendly Documentation** includes an MCP server for the docs themselves, llms.txt format, and markdown export for any documentation page.

To get started, install FastMCP using pip or uv, then create a simple server by decorating Python functions with `@mcp.tool`. Run the server with `mcp.run()` and connect clients using the FastMCP Client class. The framework integrates with popular AI SDKs and web frameworks, and offers free hosting through Prefect Horizon for deployment.

## Features
- MCP server creation with Python decorators
- MCP client for connecting to servers
- Tools, resources, and prompts components
- Multiple providers (local, filesystem, proxy, skills, custom)
- Transforms for namespacing and filtering
- Authentication and authorization support
- HTTP/SSE and WebSocket transports
- CLI for running and testing servers
- Integration with AI SDKs
- Web framework integrations
- LLM-friendly documentation
- Project configuration management
- Testing utilities

## Integrations
Prefect Horizon, OpenAPI, AI SDKs, Web frameworks

## Platforms
WINDOWS, MACOS, LINUX, API, DEVELOPER_SDK

## Pricing
Open Source

## Version
3.0.0

## Links
- Website: https://gofastmcp.com
- Documentation: https://gofastmcp.com/getting-started/welcome
- Repository: https://github.com/jlowin/fastmcp
- EveryDev.ai: https://www.everydev.ai/tools/fastmcp