ms-todo-mcp
An MCP server to read and manage your Microsoft To Do tasks through the Microsoft Graph To Do API.
README
Microsoft To Do MCP server
An MCP server that lets an LLM client (Claude Desktop, Cowork, etc.) read and manage your Microsoft To Do tasks through the Microsoft Graph To Do API.
It exposes seven tools:
| Tool | What it does | Write? |
|---|---|---|
todo_list_lists |
List your task lists | – |
todo_list_tasks |
List tasks in a list (by id or name), optional status filter | – |
todo_create_task |
Create a task (title, note, due date, importance) | ✓ |
todo_update_task |
Update title / note / due date / importance / status | ✓ |
todo_complete_task |
Mark a task completed | ✓ |
todo_delete_task |
Delete a task (permanent) | ✓ |
todo_create_list |
Create a new task list | ✓ |
How auth works
Microsoft Graph does not support application-only (daemon) permissions for creating or updating To Do tasks — only delegated permissions work. This server therefore signs in as you using the OAuth2 device-code flow and caches the resulting access + refresh tokens on disk, so you only sign in once.
The token cache contains live credentials. It is written to
~/.ms_todo_mcp/token_cache.jsonby default and is git-ignored. Never commit it.
1. Register an app in Microsoft Entra ID
- Entra admin center → App registrations → New registration.
- Name it e.g.
ms-todo-mcp. For Supported account types pick Accounts in this organizational directory only (single tenant) unless you need broader access. - No redirect URI is required for the device-code flow. Click Register.
- Authentication → Advanced settings → set Allow public client flows = Yes (required for device-code).
- API permissions → Add a permission → Microsoft Graph →
Delegated permissions → add
Tasks.ReadWrite. If your tenant requires it, click Grant admin consent. - Copy the Application (client) ID and Directory (tenant) ID from the app's Overview page.
2. Configure
cp .env.example .env
# edit .env and set MS_TODO_CLIENT_ID and MS_TODO_TENANT_ID
| Variable | Required | Notes |
|---|---|---|
MS_TODO_CLIENT_ID |
✓ | Application (client) ID |
MS_TODO_TENANT_ID |
✓ for work/school | Tenant ID/domain, or organizations / common |
MS_TODO_TOKEN_CACHE |
– | Override token-cache path |
MS_TODO_TIMEZONE |
– | IANA tz for due dates (default UTC) |
3. Install
Install the package into the Python interpreter your MCP client will launch.
By default that is plain python on your PATH. Because the client starts the
server from its own working directory, the simplest, path-free setup is to
install into that default interpreter — not a virtualenv:
pip install -e . # from the repo root (use `pip install .` for non-editable)
On Windows this also places an ms-todo-mcp launcher in your Python Scripts\
folder.
Want isolation instead of a global install? Don't reach for a bare venv with the default config —
"command": "python"won't see a venv package. Use pipx (pipx install .), which isolates the package and puts thems-todo-mcpcommand on yourPATH; then set"command": "ms-todo-mcp"in the config. (A manual venv also works, but only if the config points at that venv'spython.exeby absolute path — seeexamples/README.md.)
4. Sign in once
ms-todo-mcp login
Follow the printed instructions: open the URL, enter the code, sign in. The token cache is then stored and reused.
5. Run
ms-todo-mcp # stdio transport (for local MCP clients)
ms-todo-mcp --http # streamable HTTP on :8000 (for remote use)
6. Add it to your MCP client
For a local stdio client (e.g. Claude Desktop / Cowork custom connector), add
an entry like the following. A ready-to-edit copy lives at
examples/claude_desktop_config.json and
uses your default Python on PATH (no virtualenv path hardcoded):
{
"mcpServers": {
"ms-todo": {
"command": "python",
"args": ["-m", "ms_todo_mcp"],
"env": {
"MS_TODO_CLIENT_ID": "<your-client-id>",
"MS_TODO_TENANT_ID": "<your-tenant-id>",
"MS_TODO_TIMEZONE": "Europe/Madrid"
}
}
}
}
Notes:
- The config file lives at
%APPDATA%\Claude\claude_desktop_config.json(Windows) or~/Library/Application Support/Claude/claude_desktop_config.json(macOS). Restart the app after editing it. - For
python -m ms_todo_mcpto resolve, install the package into that default Python — runpip install -e .(orpip install .) without an active virtualenv. Ifpythonisn't your launcher, usepython3or the absolute interpreter path. On Windows you can instead set"command": "ms-todo-mcp"(the console script in your PythonScriptsfolder). - Run
ms-todo-mcp loginonce first so the token cache exists before the client launches the server (the device-code prompt goes to stderr, not chat). - Credentials live in
envhere because the repo.envis not read when the client starts the server (its working directory isn't the project).
Usage examples (what to ask your assistant)
- "List my To Do lists."
- "Show the unfinished tasks in my 'Tasks' list."
- "Add 'Send Microsoft invoices to Diego' to my Work list, due Friday, high importance."
- "Mark the task about UAT as completed."
Development
pip install -e ".[dev]"
ruff check .
python -m py_compile src/ms_todo_mcp/*.py
Project layout
ms-todo-mcp/
├── pyproject.toml
├── README.md
├── CONTRIBUTING.md
├── LICENSE # MIT
├── .env.example
├── .gitignore
├── examples/
│ ├── README.md
│ └── claude_desktop_config.json
└── src/
└── ms_todo_mcp/
├── __init__.py
├── __main__.py
├── auth.py
├── graph.py
└── server.py
Releases
Releases are automated with release-please. You do not create tags or GitHub releases by hand.
The version is derived from Conventional Commits.
Because main is squash-merged, the PR title is what counts — prefix it
accordingly:
fix: ...→ patch bump (0.1.0→0.1.1)feat: ...→ minor bump (0.1.0→0.2.0)feat!: ...or aBREAKING CHANGE:footer → major bump (0.1.0→1.0.0)docs: ...,chore: ...,ci: ...,refactor: ...→ no release
Flow:
- Merge PRs into
mainusing Conventional Commit titles. - release-please opens (and keeps updating) a release PR titled
chore: release X.Y.Z, bumping the version inpyproject.tomlandsrc/ms_todo_mcp/__init__.pyand updatingCHANGELOG.md. - When you merge that release PR, release-please creates the
vX.Y.Zgit tag and the GitHub Release with auto-generated notes.
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。