modelcontextprotocol/python-sdk

The MCP Python SDK is the official Python library for building servers and clients that use the Model Context Protocol (MCP). MCP is a standardized way for AI applications to access external data and tools, think of it like a bridge between large language models (LLMs) and external capabilities like databases, APIs, or custom functions. The problem it solves: AI assistants like Claude can only know what's in their conversation window. MCP lets developers create "servers" that feed structured information and actions into an AI, in a consistent way any MCP-compatible app can understand. This separates the job of "providing context" from the job of "talking to the AI," making both parts easier to build and maintain. How it works: you write a Python server using the FastMCP helper, decorate functions as tools (things the AI can call to do work), resources (data the AI can read), or prompts (reusable instruction templates). The server then communicates with any MCP-compatible AI client over standard transport channels like HTTP or stdio. You would use this when building an AI-powered app that needs to give an AI model access to live data or custom actions, for example, a chatbot that can look up customer records, run calculations, or fetch web content on demand. The tech stack is Python, distributed as a package installable via pip or uv.