AI-native task management for coding agents - Persistent, version-controlled TODO tracking that works naturally with AI agents like OpenClaw, Cursor, Claude, and Copilot

ai-todo

AI-native task management for coding agents

Simple, persistent, version-controlled TODO tracking that works naturally with AI agents like Cursor, Cursor, and Cursor.

Quick Start

Add this to your project's .cursor/mcp.json:

{
  "mcpServers": {
    "ai-todo": {
      "command": "uvx",
      "args": ["ai-todo", "serve", "--root", "${workspaceFolder}"]
    }
  }
}

Then enable the server in Cursor Settings → MCP Servers → toggle ai-todo on.

That's it! Your AI agent can now manage tasks directly. No installation required.

Try it: Ask your agent to "create a task for implementing user authentication"

Why ai-todo?

AI agents track tasks internally, but this creates a closed system that gets lost after sessions end. ai-todo provides a permanent, version-controlled record in your Git repository.

• Persistent — Tasks survive across sessions, restarts, and time
• Version Controlled — Tracked in Git alongside your code
• AI-Native — MCP integration for direct agent interaction
• Human Readable — Plain Markdown in standard TODO.md format
• Zero Config — Works immediately, no setup required
• Instant & Local — No API calls, authentication, or rate limits

Installation Options

Option A: Zero-Install MCP (Recommended)

For AI agent integration via Cursor or similar IDEs. Uses uvx to run on-demand without permanent installation.

Project-specific setup (.cursor/mcp.json):

{
  "mcpServers": {
    "ai-todo": {
      "command": "uvx",
      "args": ["ai-todo", "serve", "--root", "${workspaceFolder}"]
    }
  }
}

Requires uv to be installed (curl -LsSf https://astral.sh/uv/install.sh | sh).

Option B: System Installation

For CLI usage or permanent MCP server setup. Requires Python 3.10+.

# Install globally (recommended)
uv tool install ai-todo

# Or with pipx
pipx install ai-todo

CLI Usage: `ai-todo [command