onenote_mcp_server
MCP server that connects AI tools to Microsoft OneNote for Business via the Microsoft Graph API, enabling listing, creating, and updating notebooks, sections, and pages.
README
OneNote MCP Server
A Model Context Protocol (MCP) server that connects GitHub Copilot (and other MCP-compatible AI tools) to Microsoft OneNote for Business via the Microsoft Graph API.
Features
| Tool | Description |
|---|---|
list_notebooks |
List all OneNote notebooks accessible to the configured service account |
list_sections |
List sections in a notebook |
create_section |
Create a new section in a notebook |
create_page |
Create a new page in a section (HTML content) |
update_page |
Update an existing page with patch commands |
get_page |
Retrieve a page's HTML content |
Prerequisites
- Node.js 18 or later
- A Microsoft 365 / Office 365 work or school account
- An Azure AD app registration with a client secret and application permissions (one-time setup, free, takes ~5 minutes)
Step 1 — Register an Azure AD App
This is a one-time setup. You need an Azure account that has access to your organisation's Azure AD.
- Go to https://entra.microsoft.com and sign in.
- Navigate to Identity → Applications → App registrations → New registration.
- Fill in:
- Name:
OneNote MCP Server(or any name you prefer) - Supported account types: Single tenant is recommended for background-service access.
- Name:
- Click Register.
- On the app overview page, copy:
- Application (client) ID → this is your
ONENOTE_CLIENT_ID - Directory (tenant) ID → this is your
ONENOTE_TENANT_ID
- Application (client) ID → this is your
- Go to Certificates & secrets → New client secret and copy the secret value.
- Go to API permissions → Add a permission → Microsoft Graph → Application permissions.
- Add
Notes.ReadWrite.All. - Click Grant admin consent for your tenant.
Step 2 — Install the MCP Server
# Clone or download this repository, then:
cd onenote-mcp-server
npm install
npm run build
Step 3 — Use from GitHub in any project (no local clone required)
After you push this repo to GitHub, you can reference it directly from MCP host configs:
{
"servers": {
"onenote": {
"command": "npx",
"args": ["-y", "github:<YOUR_GITHUB_USERNAME>/onenote_mcp_server"],
"env": {
"ONENOTE_CLIENT_ID": "your-client-id-from-step-1",
"ONENOTE_TENANT_ID": "your-tenant-id",
"ONENOTE_CLIENT_SECRET": "your-client-secret",
"ONENOTE_USER_PRINCIPAL_NAME": "service-account@yourtenant.com"
}
}
}
}
Replace
<YOUR_GITHUB_USERNAME>/onenote_mcp_serverwith your actual GitHub repo path. Thepreparescript builds TypeScript automatically when the package is installed from GitHub.
Step 4 — Configure GitHub Copilot Desktop
Option A: Workspace configuration (recommended for team sharing)
Edit .vscode/mcp.json in this project (already provided) and fill in your values:
{
"servers": {
"onenote": {
"command": "node",
"args": ["/absolute/path/to/onenote-mcp-server/build/index.js"],
"env": {
"ONENOTE_CLIENT_ID": "your-client-id-from-step-1",
"ONENOTE_TENANT_ID": "your-tenant-id",
"ONENOTE_CLIENT_SECRET": "your-client-secret",
"ONENOTE_USER_PRINCIPAL_NAME": "service-account@yourtenant.com"
}
}
}
}
Windows path example:
"C:\\Users\\you\\onenote-mcp-server\\build\\index.js"
Option B: User-level configuration (applies to all your workspaces)
In VS Code, open the Command Palette (Ctrl+Shift+P) → MCP: Open User Configuration, then add:
{
"servers": {
"onenote": {
"command": "node",
"args": ["/absolute/path/to/onenote-mcp-server/build/index.js"],
"env": {
"ONENOTE_CLIENT_ID": "your-client-id",
"ONENOTE_TENANT_ID": "your-tenant-id"
}
}
}
}
Step 5 — Start Using the Server
The server authenticates automatically with the configured app registration and accesses OneNote for the user principal name you set in ONENOTE_USER_PRINCIPAL_NAME.
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
ONENOTE_CLIENT_ID |
✅ Yes | — | Azure AD Application (client) ID |
ONENOTE_TENANT_ID |
✅ Yes | — | Tenant ID for the Entra app registration |
ONENOTE_CLIENT_SECRET |
✅ Yes | — | Client secret value for the app registration |
ONENOTE_USER_PRINCIPAL_NAME |
✅ Yes | — | User principal name whose OneNote notebooks the service accesses |
ONENOTE_LOG_LEVEL |
No | error |
Log verbosity: debug, info, warn, error |
Usage Examples
Once configured, you can ask Copilot things like:
- "List all my OneNote notebooks"
- "Show me the sections in my Work notebook"
- "Create a section called 'Meeting Notes' in my Project notebook"
- "Create a OneNote page titled 'Sprint Planning' with a bullet list of our team goals"
- "Get the content of page [ID] so I can update it"
- "Add a new paragraph to the end of page [ID]"
Token Cache Location
After the service first authenticates, the token cache is stored here (never committed to git):
| Platform | Location |
|---|---|
| Windows | %APPDATA%\onenote-mcp-server\token-cache.json |
| macOS | ~/Library/Application Support/onenote-mcp-server/token-cache.json |
| Linux | ~/.local/share/onenote-mcp-server/token-cache.json |
The cache file is restricted to owner-only permissions (chmod 600 on Unix).
To force re-authentication, delete this file.
Security
- Confidential client app — client secrets are required and kept in environment variables
- Application permissions —
Notes.ReadWrite.Allvia the Graph.defaultscope - Client credentials flow — the server authenticates headlessly in the background
- Tokens never logged — the
Authorizationheader and token values are never written to logs - Input validation — all resource IDs are validated against a safe-ID pattern before URL interpolation
- HTML sanitisation — script tags, event handlers, and dangerous URI schemes are stripped from page content
- Atomic cache writes — token cache is written via temp file + rename to prevent corruption
Scopes and Permissions
The server requests Notes.ReadWrite.All as an application permission, which covers:
- Reading notebooks, sections, and pages for the configured user principal name
- Creating notebooks, sections, and pages
- Updating pages
Admin consent is required for the app permission.
Troubleshooting
"ONENOTE_CLIENT_ID environment variable is not set"
→ Make sure ONENOTE_CLIENT_ID is set in your mcp.json env block.
"ONENOTE_CLIENT_SECRET environment variable is not set"
→ Make sure the client secret value is set in your mcp.json env block.
"ONENOTE_USER_PRINCIPAL_NAME environment variable is not set"
→ Make sure the service account UPN is set in your mcp.json env block.
"Graph API error: 403"
→ The app registration is missing Notes.ReadWrite.All, admin consent was not granted, or the configured user principal name cannot access the requested notebook.
"Graph API error: 401"
→ Delete the token cache file (listed above under "Token Cache Location") and re-authenticate.
"Request timed out"
→ Network issue connecting to graph.microsoft.com. Check your proxy/firewall settings.
Use with other MCP hosts (Claude Desktop / Cursor / others)
Any MCP host that supports stdio can use this server with the same command/env pattern:
{
"mcpServers": {
"onenote": {
"command": "npx",
"args": ["-y", "github:<YOUR_GITHUB_USERNAME>/onenote_mcp_server"],
"env": {
"ONENOTE_CLIENT_ID": "your-client-id",
"ONENOTE_TENANT_ID": "your-tenant-id",
"ONENOTE_CLIENT_SECRET": "your-client-secret",
"ONENOTE_USER_PRINCIPAL_NAME": "service-account@yourtenant.com"
}
}
}
}
If your host expects servers instead of mcpServers, keep the same inner object and rename only the top-level key.
Development
# Run without building (uses tsx)
npm run dev
# Build TypeScript
npm run build
# Start the built server
npm start
API Reference
All operations use the Microsoft Graph API v1.0 endpoints:
| Operation | Endpoint |
|---|---|
| List notebooks | GET /users/{userPrincipalName}/onenote/notebooks |
| List sections | GET /users/{userPrincipalName}/onenote/notebooks/{id}/sections |
| Create section | POST /users/{userPrincipalName}/onenote/notebooks/{id}/sections |
| Create page | POST /users/{userPrincipalName}/onenote/sections/{id}/pages |
| Update page | PATCH /users/{userPrincipalName}/onenote/pages/{id}/content |
| Get page | GET /users/{userPrincipalName}/onenote/pages/{id}/content |
Note: The server uses application permissions to access OneNote for the configured user principal name. Admin consent is required.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。