microsoft/mcp

What it solves

This project provides a standardized way for AI applications to access data and tools from various sources. It implements the Model Context Protocol (MCP), which eliminates the need to build custom, divergent connectors for every different AI assistant or IDE.

How it works

It uses a client-server architecture consisting of three components:

• MCP Hosts: The main applications (like IDEs or AI assistants) that start the connection.
• MCP Clients: Connectors within the host that maintain a 1:1 link with a server.
• MCP Servers: Services that provide the actual context and capabilities to the LLM.

This repository specifically contains the core libraries, test frameworks, and tooling used to build Microsoft's MCP servers, such as those for Azure, Microsoft Fabric, and various Microsoft 365 services.

Who it’s for

• AI Developers: Those building AI agents or assistants that need to connect to cloud infrastructure, developer tools, or productivity software.
• MCP Server Contributors: Developers looking to unify engineering investments and reduce duplication when creating new MCP servers.

Highlights

• Broad Integration: Includes servers for Azure, GitHub, Azure DevOps, AKS, and Microsoft 365 (Mail, Calendar, Copilot Chat).
• Unified Engineering: Provides shared libraries and pipelines to prevent divergence across different server implementations.
• Versatile Deployment: Supports both local servers (via stdio) and remote servers (via HTTP).
• IDE Integration: Deeply integrated with VS Code, Visual Studio, and IntelliJ IDEA.