MCP 服务器

OneNote-MCP-For-GeminiSpark

A robust MCP server that transforms OneNote notebooks into an AI-accessible knowledge base for Gemini Spark, enabling natural language queries to list, read, and search notes via Microsoft Graph API.

README

OneNote-MCP-For-GeminiSpark

A robust Model Context Protocol (MCP) server for Microsoft OneNote integration with Gemini Spark. This repository contains code, configuration, and examples to synchronize OneNote content with GeminiSpark workflows, export OneNote notes into GeminiSpark-ready formats, and demonstrate common automation scenarios.

🎯 What This Does

Transform your OneNote notebooks into an AI-accessible knowledge base:

List all your notebooks, sections, and pages
Read page content for analysis and search
Natural language queries like "Show me my DevOps notes" or "Find pages about project planning"
Secure OAuth authentication with Microsoft Graph API
Bulletproof error handling with detailed debugging

✨ Why This Implementation

Unlike other OneNote MCP servers, this one:

✅ Actually works - tested extensively with real OneNote data
✅ Complete functionality - all core OneNote operations implemented
✅ Robust authentication - two-step device flow that handles edge cases
✅ Production ready - proper error handling and logging
✅ Easy setup - detailed instructions for non-technical users

🚀 Quick Start

Prerequisites

Python 3.10+
uv package manager (recommended) or pip
Google Gemini Account
Microsoft Azure account (free)

1. Install uv (if you don't have it)

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# or with Homebrew
brew install uv

2. Clone and Setup

git clone https://github.com/yourusername/onenote-mcp-server.git
cd onenote-mcp-server

# Create virtual environment and install dependencies
uv sync

3. Azure App Registration

You need to create an Azure app to access OneNote. Don't worry, it's free and takes 5 minutes:

Go to Azure Portal (sign in with your Microsoft account)
Navigate to Azure Active Directory → App registrations → New registration
Fill out the form:
- Name: "OneNote MCP Server" (or whatever you like)
- Supported account types: "Accounts in any organizational directory and personal Microsoft accounts"
Click Register
Copy the Application (client) ID - you'll need this!

4. Add Permissions

Still in your Azure app:

Go to API permissions → Add a permission
Select Microsoft Graph → Delegated permissions
Add these permissions:
- Notes.Read - Read OneNote notebooks
- Notes.ReadWrite - Create/modify OneNote content (optional but recommended)
- User.Read - Read user profile
- Offline.access - Permit Refresh Token Access
Click Grant admin consent (the button at the top)
Go to Manifest -> Update signInAudience to AzureADandPersonalMicrosoftAccount and requestedAccessTokenVersion to 2 and Save.

5. Generate Azure Refresh Token (One Time Activity)

Update Azure Client ID in refresh_token.py file and run it on cmd.
Browser window will open. Login with your Microsoft ID.
Post successful login, Token will be generated in cmd.

6. Deploy MCP Server on Cloud/Locally

e.g. for Google Cloud Deployment run following command.

gcloud run deploy onenote-mcp-server --source . --region us-central1
--set-env-vars="AZURE_REFRESH_TOKEN=YOUR_AZURE_REFRESH_TOKEN,AZURE_CLIENT_ID=YOUR_ACTUAL_CLIENT_ID"
--allow-unauthenticated

Post Deployment MCP Server URL will be generated.

7. Setup Custom Connected App in Gemini Spark

Click on Connected Apps
Fill MCP Server URL/sse (Do not forget to add /sse at the of the URL)
In Advance Settings -> Fill Azure Client ID
Click on Copy Redirect URI and Go to Azure Portal --> Azure Client App --> Authentication --> Redirect URIs --> Add URI --> Web --> Paste the Copied Link --> Remove any other Redirect URIs
On the Connected App Dialog box -> Click next and Follow screens.
Onenote Mcp server will be created in Connected App Screen

Usage

Once Connected App is setup, try these tasks in Gemini Spark :

List my OneNote notebooks
Show me sections in my Work notebook  
What pages are in my Ideas section?
Read the content of my "Project Plan" page

🛠 Troubleshooting

"No tools available" in Gemini Spark Connected Apps Screen

Make sure you copied redirect URI from Screen and updated on Microsoft Azure App correctly.
Check if your cloud deployment of MCP server is enabled for unauthenticated access.
Verify uv is installed: uv --version

Authentication issues

Safari OAuth problems: Safari may not handle Microsoft's OAuth redirect properly - use Firefox or Chrome instead
"nativeclient" prompts: Normal Microsoft OAuth behavior, but if it blocks authentication, try a different browser
Recommended browsers: Firefox (confirmed working), Chrome, or Edge for best compatibility

"Command not found" errors

Make sure uv is in your PATH
Alternative: replace "uv" with "python" in the config and use the full path to your Python interpreter

Permission denied errors

Check the file permissions in your project directory

🏗 Development

Project Structure

onenote-mcp-server/
├── onenote_mcp_server.py      # Main server implementation
├── pyproject.toml             # Dependencies and metadata
├── refresh_token.py           # Generate Refresh Token
├── README.md                  # This file
├── LICENSE                    # MIT License
└── .gitignore                 # Git ignore rules

Key Features

Complete Graph API integration: All OneNote operations supported
Robust error handling: Detailed logging and graceful failures
FastMCP framework: Clean, maintainable code structure
Environment variable configuration: Secure credential handling

Adding New Features

The server is built with FastMCP, making it easy to add new tools:

if method == "tools/list":
                    await custom_send({"type": "http.response.start", "status": 200, "headers": [(b"content-type", b"application/json")]})
                    tools = [
  
    # Your implementation here
    return