prefecthq/fastmcp

FastMCP is a Python framework for building MCP servers and clients. MCP stands for Model Context Protocol, an open standard that lets AI language models (LLMs) connect to external tools and data sources. Think of it as the "USB port" that lets an AI assistant plug in to your code, your database, or any custom function you write. FastMCP makes building those connections fast and Pythonic (meaning natural and clean to write in Python). The problem it solves: connecting LLMs to your own tools without writing a lot of boilerplate. You decorate a Python function with a simple @mcp.tool tag, and FastMCP automatically handles the schema generation, input validation, and protocol communication. You focus on what your tool does; FastMCP handles the AI-plumbing. It has three main parts. Servers let you expose your Python functions as tools that an AI can call. Clients let you connect from your code to any MCP server. Apps let your tools have interactive UIs that render directly inside an AI conversation. FastMCP 1.0 was incorporated into the official MCP Python SDK in 2024, and the README states it powers 70% of MCP servers across all languages. You'd use this when building an AI agent or chatbot that needs to call real-world functions, querying a database, running calculations, reading files, or calling an API. It's ideal for developers who want to give an LLM access to their own backend logic without deep protocol knowledge. Built in Python by Prefect.