A Model Context Protocol (MCP) server implementation that enables AI language models like Claude and other LLMs to interact with Microsoft OneNote.
This project is based on azure-onenote-mcp-server by Zubeid Hendricks, with modifications to simplify authentication and improve usability.
This server allows AI assistants to:
- Access your OneNote notebooks, sections, and pages
- Create new pages in your notebooks
- Search through your notes
- Read complete note content, including HTML formatting and text
- Analyze and summarize your notes directly
All of this happens directly through the AI interface without you having to switch contexts.
-
Clone this repository and follow the installation steps below
-
Start the MCP server:
npm start -
Register the server in Cursor:
- Open Cursor preferences (Cmd+, on Mac or Ctrl+, on Windows)
- Go to the "MCP" tab
- Add a new MCP server with these settings:
- Name:
onenote - Command:
node - Args:
["/path/to/your/onenote-mcp.mjs"](use absolute path)
- Name:
Here's the complete JSON configuration example:
{ "mcpServers": { "onenote": { "command": "node", "args": ["/absolute/path/to/your/onenote-mcp.mjs"], "env": {} } } } -
Restart Cursor
-
In Cursor, you can now interact with your OneNote data using natural language:
Can you show me my OneNote notebooks?
Create a new page in my first notebook with a summary of this conversation
Find notes related to "project planning" in my OneNote
The first time you ask about OneNote, the AI will guide you through the authentication process.