Access and Manage OneNote Data

OneNote MCP Server is an MCP server that gives an AI assistant access to Microsoft OneNote, reading and writing notebooks, sections, and pages through the Microsoft Graph API. It also converts HTML content to plain text, useful for feeding OneNote pages into RAG pipelines more cleanly.

Use it when you want an AI assistant to read existing OneNote content or create new notebooks, sections, or pages, with HTML-formatted content. It requires a Microsoft Azure account with a registered application, a CLIENT_ID, and the appropriate Graph API permissions, Notes.Read and Notes.ReadWrite, so it depends on Azure app registration being done first. Authentication uses device code flow: on first run, a code is generated and written to device-code.txt, which you enter at the URL Microsoft provides, after which tokens are cached in token-cache.json so you don't have to repeat the flow on every run.

onenote-read reads content from OneNote notebooks, sections, or pages, with optional metadata and content details. onenote-create creates new content in OneNote, including pages, sections, or notebooks, with HTML content. Beyond the two tools, the server handles HTML-to-text conversion for cleaner downstream processing, and caches authentication tokens so repeated runs don't require re-authenticating via device code each time.

From source:

git clone [repository]
npm install
npm run build
npm start

Or via Docker:

mkdir -p data
docker build -t onenote-mcp-server .
docker run -d --name onenote-mcp-server -e CLIENT_ID=your-client-id -v $(pwd)/data:/app/dist onenote-mcp-server

For Claude Desktop, set the server directory to your cloned repository, the command to npm run build && npm start, and add the CLIENT_ID environment variable with your Microsoft Azure Application Client ID.

Developers who want an AI assistant to read from or write to Microsoft OneNote directly, provided they're willing to register an Azure application and complete a one-time device code authentication flow.