Microsoft Graph MCP Server

A beautifully simple way to give your AI assistant superpowers inside Microsoft 365.

This project turns Microsoft Graph into an easy‑to‑use MCP server — so Claude, your custom agents, or any MCP client can instantly work with your Email, Calendar, and Contacts.

No Azure setup. No Graph API knowledge. Just install → sign in → your AI becomes a Microsoft 365 power user.

---

⭐ Why Developers Love This Project

• Your AI can read, search, reply to, and organize email naturally
• Batch email operations — delete or archive multiple emails in ONE call (no more one-by-one!)
• It understands and manages your calendar with real availability checks
• It handles contacts with simple MCP tools
• It works securely using Microsoft’s device login — no secrets, no risk
• It just works on local machines with zero configuration

This is the fastest way to make your AI agent actually useful at work.

If this project saves you time:
👉 Please star the repo — it really helps! ⭐

---

🚀 Quick Start

1. Install

pip install -r requirements.txt

Multimodal Support: Image compression requires Pillow>=10.0.0 (included in requirements.txt). If not installed, images will be returned without compression.

Or:

pip install uv
uvx .

2. Link to Claude Desktop

Open:

%APPDATA%/Claude/claude_desktop_config.json

Choose one of the following configuration methods:

Method 1: Using uv run (Recommended)

{
  "mcpServers": {
    "microsoft-graph": {
      "command": "uv",
      "args": ["run", "--directory", "", "microsoft-graph-mcp-server"]
    }
  }
}

Recommended: This method uses your local code directly, so changes take effect immediately without reinstalling.

Method 2: Using uvx (Local Path)

{
  "mcpServers": {
    "microsoft-graph": {
      "command": "uvx",
      "args": ["--from", "", "microsoft-graph-mcp-server"]
    }
  }
}

Note:

• First time or after code changes, run: uv tool install --force
• uvx caches the package, so you need to reinstall to pick up code updates

Method 3: Using python -m

{
  "mcpServers": {
    "microsoft-graph": {
      "command": "py",
      "args": ["-m", "microsoft_graph_mcp_server"]
    }
  }
}

Note:

• Windows: Use py (Python Launcher) which automatically finds your Python installation
• Linux/Mac: Replace py with python or python3
• Run from the project directory or install via pip install -e .

3. Sign In

In Claude, simply ask: “Please help me sign in to Microsoft Graph”

Claude will call:

{
  "tool": "auth",
  "action": "start"
}