FastMCP Documentation
by Prefect (FastMCP)
The production reference for building MCP servers, clients and apps in Python.
Overview
FastMCP is a full framework for building Model Context Protocol applications in Python, maintained by the Prefect team after the project moved from jlowin/fastmcp to PrefectHQ/fastmcp with the 3.0 release on 18 February 2026. The docs are organised around three pillars — Servers, Clients and Apps — plus Integrations, Deployment, a CLI and development section, and a full Python SDK reference. The Servers track starts with exposing plain Python functions as tools, resources and prompts, with schemas generated from ordinary type hints, then goes deep: tool transforms including code mode and tool search, component providers (local, filesystem, proxy, custom), middleware, dependency injection, lifespans, storage, logging, progress reporting, pagination and telemetry. Security gets its own run of pages covering authentication patterns, OAuth and authorization, with concrete integrations for Auth0, AuthKit, AWS Cognito, Azure, GitHub, Google, Keycloak, Supabase and WorkOS. The Clients track covers calling tools, reading resources and fetching prompts programmatically, with OAuth, machine-to-machine, bearer-token and CIMD auth. Integration pages wire servers into FastAPI, OpenAPI, ChatGPT, Claude Desktop and Claude Code, Cursor, Gemini CLI and Goose, and into the Anthropic, OpenAI, Gemini and Pydantic AI SDKs. Deployment covers local runs, HTTP, sandboxed agents and fastmcp.json project config. Version 3.4.5 shipped 27 July 2026 as the current stable line, with 4.0.0b1 in beta from 28 July 2026.
At a Glance
- Topic
- MCP
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Reference docs; ~4-6 hours to work through the server and client tracks, self-paced
- Provider
- Prefect (FastMCP)
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Expose ordinary Python functions as MCP tools, resources and prompts
- ✓Generate and validate MCP tool schemas directly from Python type hints
- ✓Write MCP clients that call tools, read resources and fetch prompts
- ✓Secure a server with OAuth, bearer tokens or machine-to-machine credentials
- ✓Compose, proxy and transform existing servers instead of rewriting tool code
- ✓Generate an MCP server automatically from an existing FastAPI or OpenAPI app
- ✓Deploy over HTTP and pin project configuration in a fastmcp.json file
- ✓Test and inspect a running server using the bundled fastmcp CLI
Highlights
- •FastMCP 1.0 was folded into the official MCP Python SDK, so these idioms are the reference ones rather than a third-party dialect
- •27.2k GitHub stars, Apache-2.0 licensed, with full-time engineering support from the Prefect team behind it
- •Covers the unglamorous production surface most MCP tutorials skip entirely: auth, middleware, telemetry, pagination and deployment
- •Integration pages for Claude Code, Cursor, ChatGPT, Gemini CLI and Goose mean you can test against a real host the same day
- •Caveat worth knowing: the site tracks the main branch, so v4 beta material sits next to the stable 3.x line — check the version note before copying code
Who It’s For
Best For
- ✓Python engineers shipping MCP servers to production
- ✓Platform teams standardising internal tool APIs for LLM agents
- ✓Developers migrating off the low-level MCP Python SDK
- ✓Teams that need OAuth-protected remote MCP servers
Prerequisites
- •Intermediate Python, including async/await and type hints
- •Familiarity with MCP concepts: tools, resources, prompts and hosts
- •A working Python 3 environment with uv or pip
FAQ
What is FastMCP Documentation?
The official documentation for FastMCP, the Python framework that turns ordinary functions into Model Context Protocol tools, resources and prompts. Written for engineers who need more than a hello-world server: it covers authentication, middleware, proxying, telemetry and deployment. Work through it and you can ship an OAuth-protected remote MCP server that real hosts like Claude Code and Cursor can connect to.
Is FastMCP Documentation free?
FastMCP Documentation is free to access.
What level is FastMCP Documentation for?
FastMCP Documentation is aimed at a intermediate audience. Recommended background: Intermediate Python, including async/await and type hints, Familiarity with MCP concepts: tools, resources, prompts and hosts, A working Python 3 environment with uv or pip.
How long does FastMCP Documentation take?
Expect roughly Reference docs; ~4-6 hours to work through the server and client tracks, self-paced. Most learners work through it at their own pace.
What will I learn from FastMCP Documentation?
You'll learn: Expose ordinary Python functions as MCP tools, resources and prompts; Generate and validate MCP tool schemas directly from Python type hints; Write MCP clients that call tools, read resources and fetch prompts; Secure a server with OAuth, bearer tokens or machine-to-machine credentials; Compose, proxy and transform existing servers instead of rewriting tool code; Generate an MCP server automatically from an existing FastAPI or OpenAPI app; Deploy over HTTP and pin project configuration in a fastmcp.json file; Test and inspect a running server using the bundled fastmcp CLI.
Topics
Sources
This page was written from 4 sources, 2 on domains other than gofastmcp.com.