TaskWarrior MCP Server
# TaskWarrior MCP Server
Node.js server implementing Model Context Protocol (MCP) for [TaskWarrior](https://taskwarrior.org/) operations.
<a href="https://glama.ai/mcp/servers/e8w3e1su1x">
<img width="380" height="200" src="https://glama.ai/mcp/servers/e8w3e1su1x/badge" alt="TaskWarrior Server MCP server" />
</a>
## Features
- View pending tasks
- Filter tasks by project and tags
- Add new tasks with descriptions, due dates, priorities, projects and tags
- Mark tasks as complete
**Note**: This runs your local `task` binary, so TaskWarrior needs to be installed and configured!
> [!WARNING]
> This currently uses task `id` which is an unstable identifier; taskwarrior
> sometimes renumbers tasks when new ones are added or removed. In the future
> this should be more careful, using task UUID instead.
## API
### Tools
- **get_next_tasks**
- Get a list of all pending tasks
- Optional filters:
- `project`: Filter by project name
- `tags`: Filter by one or more tags
- **add_task**
- Add a new task to TaskWarrior
- Required:
- `description`: Task description text
- Optional:
- `due`: Due date (ISO timestamp)
- `priority`: Priority level ("H", "M", or "L")
- `project`: Project name (lowercase with dots)
- `tags`: Array of tags (lowercase)
- **mark_task_done**
- Mark a task as completed
- Required:
- `identifier`: Task ID or UUID
## Usage with Claude Desktop
Add this to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"taskwarrior": {
"command": "npx",
"args": [
"-y",
"mcp-server-taskwarrior"
]
}
}
}
```
## Installation
```bash
npm install -g mcp-server-taskwarrior
```
Make sure you have TaskWarrior (`task`) installed and configured on your system.
## Example usage ideas:
* What are my current work tasks?
* Executes: `task project:work next`
* TODO: Call my sister (high priority)
* Executes: `task add priority:H Call my sister`
* OK, I've called my sister
* Executes: `task done 1`
## License
This MCP server is licensed under the MIT License. See the LICENSE file for details.TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: add_task creates new tasks, get_next_tasks retrieves pending tasks, and mark_task_done updates task status to completed. There is no overlap or ambiguity between these three core operations.
All tools follow a consistent verb_noun pattern (add_task, get_next_tasks, mark_task_done) with clear action verbs and descriptive nouns. The naming is predictable and readable throughout.
With only 3 tools, the server feels thin for task management. While the tools cover basic operations, typical task management systems benefit from more functionality like updating tasks, deleting tasks, or filtering/searching. The count is borderline minimal for the domain.
The tools provide basic CRUD coverage for tasks (create, read, update via marking done), but there are notable gaps. Missing operations include updating task details (beyond marking done), deleting tasks, and filtering/searching tasks beyond 'next' tasks. This limits agent flexibility in handling complex task workflows.