MCP OneNote Server

MCP OneNote Server

What is MCP OneNote Server about?

Enables AI models to interact with Microsoft OneNote, allowing them to retrieve content from notebooks, sections, and pages and to create or update pages programmatically.

How to use MCP OneNote Server?

1. Clone the repository https://github.com/rajvirtual/MCP-Servers.
2. Install dependencies with npm or yarn (npm install or yarn).
3. Start the server using the provided entry point (e.g., npm run start or npx -y mcp-servers).
4. Configure authentication credentials (Microsoft Graph OAuth token) in environment variables.
5. Send MCP‑formatted requests to the server’s HTTP endpoint to read or write OneNote data.

Key features of MCP OneNote Server

• Full read/write access to OneNote notebooks, sections, and pages.
• TypeScript implementation for strong typing and developer friendliness.
• Compatible with Claude‑AI and other LLMs that understand the Model Context Protocol.
• Simple HTTP endpoint that can be invoked from any AI agent runtime.
• Extensible architecture allowing additional OneNote operations to be added.

Use cases of MCP OneNote Server

• AI‑driven personal assistants that can retrieve meeting notes or add summaries to OneNote.
• Automated documentation tools that generate or update project pages.
• Knowledge‑base bots that query OneNote for answers and write new entries based on interactions.
• Workflow automation where AI decides to create task pages after processing emails or chat messages.

FAQ from the MCP OneNote Server

Q: Which authentication method is required? A: The server uses Microsoft Graph API; you need a valid OAuth access token with Notes.ReadWrite.All permission.

Q: Can the server run on a serverless platform? A: Yes, because it is a standard Node.js HTTP service, it can be deployed to AWS Lambda, Azure Functions, or similar environments.

Q: Is there support for other Microsoft 365 services? A: The current implementation focuses on OneNote. Additional services can be integrated by extending the MCP handlers.

Q: How do I add custom logic for page formatting? A: Modify the TypeScript handlers in the src directory; the server exposes hooks for preprocessing content before writing.

Q: What LLMs are compatible? A: Any model that can generate or parse MCP payloads, such as Claude, GPT‑4, or hosted LLMs that support custom protocols.

OneNote MCP Server

A Model Context Protocol (MCP) server that provides AI assistants with access to Microsoft OneNote. This server enables AI models to read from and write to OneNote notebooks, sections, and pages.

Project Overview

This project implements an MCP server that connects to Microsoft OneNote using the Microsoft Graph API. It provides tools for:

• Reading notebooks, sections, and pages from OneNote
• Creating new notebooks, sections, and pages in OneNote
• Converting HTML content to text for better RAG processing

Project Structure

onenote/
├── dist/                # Compiled JavaScript files (generated)
├── src/                 # TypeScript source files
│   ├── index.ts         # Main entry point and server implementation
│   └── types/           # Custom TypeScript type definitions
├── .vscode/             # VS Code configuration
│   ├── launch.json      # Debug configurations
│   └── tasks.json       # Build tasks
├── package.json         # Project dependencies and scripts
├── tsconfig.json        # TypeScript configuration
├── Dockerfile           # Docker configuration
├── .env.local.example   # Example environment variables
└── README.md            # This file

Authentication

The server uses Microsoft Authentication Library (MSAL) with device code flow for authentication:

1. When first run, the server generates a device code and URL
2. The code is saved to device-code.txt in the project directory
3. You must visit the URL and enter the code to authenticate
4. After authentication, tokens are cached in token-cache.json for future use

MCP Tools

The server provides the following MCP tools:

onenote-read

Read content from Microsoft OneNote notebooks, sections, or pages.

Parameters:

• type: "read_content"
• pageId: (optional) ID of the specific page to read
• sectionId: (optional) ID of the section to list pages from
• notebookId: (optional) ID of the notebook to list sections from
• includeContent: (optional) Whether to include the content of the page (default: true)
• includeMetadata: (optional) Whether to include metadata about the page (default: false)

onenote-create

Create new content in Microsoft OneNote.

Parameters:

• type: "create_page", "create_section", or "create_notebook"
• title: Title of the content to create
• content: Content in HTML format (for pages)
• parentId: (optional) ID of the parent section or notebook

Getting Started

Prerequisites

• Node.js (v14 or higher)
• npm (v6 or higher)
• Microsoft Azure account with a registered application
• OneNote account (Microsoft 365 subscription)

Azure Setup

1. Register a new application in the Azure Portal
2. Add the following API permissions:
   • Microsoft Graph > Notes.Read
   • Microsoft Graph > Notes.ReadWrite
3. Configure authentication:
   • Under "Authentication" settings, set "Supported account types" to:
     • "Accounts in any organizational directory (Any Microsoft Entra ID tenant - Multitenant) and personal Microsoft accounts (e.g. Skype, Xbox)"
   • Enable "Allow public client flows" for the app
4. Note your Application (client) ID for configuration

Installation

1. Clone the repository
2. Install dependencies:

npm install

3. Create a .env.local file with your Azure client ID:

CLIENT_ID=your-client-id-from-azure

Development

To run the application in development mode:

npm run dev

Building

To compile TypeScript to JavaScript:

npm run build

Running

To run the compiled application:

npm start

Docker Support

You can build and run the application using Docker:

# Create a data directory for persistence
mkdir -p data

# Build the Docker image
docker build -t onenote-mcp-server .

# Run the container
docker run -d \
  --name onenote-mcp-server \
  -e CLIENT_ID=your-client-id \
  -v $(pwd)/data:/app/dist \
  onenote-mcp-server

Authentication with Docker

When running in Docker, the authentication flow works as follows:

1. Start the container as shown above
2. Check the device code file:

   cat data/device-code.txt
   

3. Follow the instructions to authenticate with Microsoft
4. The token will be cached in data/token-cache.json for future use

Setup with Claude Desktop

1. Clone this repository

2. Run npm install to install dependencies

3. In Claude Desktop, add a new MCP server:

   • Set the server directory to your cloned repository
   • Set the command to: npm run build && npm start
   • Add the following environment variable:
     • Name: CLIENT_ID
     • Value: [Your Microsoft Azure Application Client ID]

4. Save the configuration and connect to the server.

   Find the file location (claude_desktop_config.json):

   The file is typically located at:

   On Mac OS (~/Library/Application Support/Claude/claude_desktop_config.json)

   On Windows (C:\Users\YourUsername\AppData\Roaming\Claude\claude_desktop_config.json)

   

Authentication Flow

1. On first run, the server will generate a device code and URL
2. The code is saved to device-code.txt in the project directory
3. Visit the URL and enter the code to authenticate
4. After authentication, tokens are cached in token-cache.json
5. Subsequent runs will use the cached token if valid

Troubleshooting

• Authentication Issues: Delete token-cache.json to force re-authentication
• Module Errors: Ensure you're using Node.js 14+ with ES modules support
• TypeScript Errors: Run npm run build to check for compilation errors