MicrosoftDocs/mcp

Official Microsoft Learn MCP Server and CLI tool – powering LLMs and AI agents with real-time, trusted Microsoft docs & code samples.

What it solves

This project provides a way for AI assistants to access the most current official Microsoft documentation and code samples, reducing "hallucinations" where AI models invent non-existent SDK methods or library packages based on outdated training data.

How it works

It implements a server based on the Model Context Protocol (MCP), allowing compatible AI clients (like VS Code, Claude, Cursor, and GitHub Copilot) to connect to a remote endpoint. The server provides three primary tools:

  • microsoft_docs_search: Performs semantic search across official technical documentation.
  • microsoft_docs_fetch: Retrieves specific documentation pages and converts them to Markdown format.
  • microsoft_code_sample_search: Searches for official Microsoft and Azure code snippets and examples.

Additionally, it offers a CLI tool (@microsoft/learn-cli) for terminal access and "Agent Skills" (instruction packages) to help agents understand when and how to use these tools effectively.

Who it’s for

Developers using AI-powered IDEs or agents to build applications with Microsoft technologies (such as .NET, Azure, C#, and F#) who need accurate, up-to-date technical information.

Highlights

  • Zero Configuration: No API keys, logins, or sign-ups are required for installation.
  • Broad Compatibility: Supports a wide range of MCP-compliant clients including VS Code, Cursor, Claude Desktop, and ChatGPT.
  • Trusted Source: Accesses only official 1st-party Microsoft documentation to ensure security and reliability.
  • Experimental Features: Includes an OpenAI-compatible endpoint for Deep Research models and token budget control to manage costs.