Taskdog MCP
The MCP server of the local Taskdog task manager: an agent creates tasks, runs their lifecycle and splits big ones into subtasks.
- 361 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent creates, changes and deletes tasks in a local database through the server's HTTP API.
Install
- Works in
Manual install
make install-mcpThe first command in the Installation section of the `taskdog-mcp` README, run from the clone root. `uv tool install taskdog-mcp` installs it globally. A running `taskdog-server` is required.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent creates, changes and deletes tasks in a local database through the server's HTTP API.
This is third-party code. Review the repository files before installing.
What it does
Taskdog is a task manager with CLI, TUI and REST API interfaces, schedule calculation and time tracking. Data lives in a local SQLite database. The `taskdog-mcp` package connects an MCP client to it, and the client works with tasks in plain phrases. The tools cover creating and editing (`create_task`, `update_task`, `delete_task`), the lifecycle (`start_task`, `complete_task`, `pause_task`, `cancel_task`, `reopen_task`), statistics, dependencies, tags and notes. `decompose_task` splits a large task into subtasks, and `get_executable_tasks` shows what the agent can do itself.
Who it is for. People who keep personal tasks in the terminal and want to talk to Claude Desktop or another MCP client about plans without opening the CLI.
Good fit when
- You dictate the day's plans and the agent creates tasks and sets priorities.
- A large task is split into subtasks and linked with dependencies.
- The agent asks which tasks it can do itself and starts them.
- Task notes are updated as work goes on, and tag statistics help summarize the week.
Not a fit when
- Team work on a shared list: the project is designed for individual use.
- The `taskdog-server` cannot be kept running: MCP reaches it over HTTP.
- A cloud service with cross-device sync is needed: data is stored locally.
Example request
Create a task "write unit tests for the API" with priority 10, decompose it into subtasks and show which of them can start today.MCP
- Transport
- stdio
- Authentication
- key is optional
Environment variables
`TASKDOG_API_BASE_URL`- The `taskdog-server` address, `http://127.0.0.1:8000` by default.
`TASKDOG_API_KEY`secret- The API key, needed when server authentication is enabled.
`TASKDOG_MCP_NAME`- The MCP server name, `taskdog` by default.
`TASKDOG_MCP_LOG_LEVEL`- The log level, `INFO` by default.
Limitations
The MCP package does not work on its own: `taskdog-server` must be running, by default at `http://127.0.0.1:8000`. Python 3.13 or newer is required. The project is designed for a single user and stores data in a local SQLite database. Linux and macOS are fully supported, Windows is still experimental and WSL2 is recommended. The `delete_task` and `update_task` tools change data, so give the agent only the tasks it needs. The API key is optional and needed only when the server requires authentication.
How to disable. Remove the `taskdog` block from the MCP client configuration and restart the client. Stop and delete `taskdog-server` and the SQLite database separately if they are no longer needed.
FAQ
How do I set the server address?
Create `~/.config/taskdog/mcp.toml` with `base_url` under `[api]` or set `TASKDOG_API_BASE_URL`. Environment variables override the file.
Is an API key required?
Only if the server requires authentication. Put the key in `api_key` of the settings file or in `TASKDOG_API_KEY`.
How do I connect the server to Claude Desktop?
Add an `mcpServers` block with a `taskdog` entry and the command `taskdog-mcp` to `claude_desktop_config.json`. A development variant uses `uv --directory ... run taskdog-mcp`.
What does decompose_task do?
It splits a large task into subtasks with AI. Dependencies between them are added with `add_dependency`.
What if I see "Cannot connect to API server"?
Check that `taskdog-server` is running: `curl http://127.0.0.1:8000/health`. Then compare the address in `mcp.toml`.
Discussion
No questions or reviews yet.
Related
MCP serversA task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI
An MCP server for Todoist: Claude creates, finds, edits, completes and deletes tasks from an ordinary phrase, with five tools
The official Todoist MCP server: the agent finds, creates and updates tasks and projects in your account
Doska kanban MCP server: an agent creates cards, moves them across columns, sets deadlines and searches every board
Microsoft To Do MCP gives an agent tasks with checklist items and recurrence rules for working on the user’s own plans.
MCP server for Super Productivity: an agent creates tasks, sorts them by project and tag, starts the timer and reads the worklog.
README in short
The repository is a monorepo of five packages: core, HTTP API client, FastAPI server, CLI and TUI, and `taskdog-mcp`. A Docker demo image with sample tasks helps to get acquainted.