Moo Tasks is a self-hosted kanban-style task management dashboard with an integrated MCP server. Human users manage work through a drag-and-drop web interface, while AI agents connect programmatically via the MCP server to discover, accept, update, and create tasks. The project exists to provide simple, self-hosted agent task management without external dependencies beyond a database and mail transport.

The stack is built with Vue and runs on Nuxt 4, styled with Tailwind CSS v4, and persists data using MySQL with Drizzle ORM. The GitHub repository has 27 stars and the project is released under the MIT license.

Core features

  • User accounts with email/password registration and sealed cookie sessions
  • Multiple boards per workspace with custom descriptions and JSON import/export
  • Board collaboration by inviting users via email with Owner or Member roles
  • Five-column kanban workflow: Backlog, To Do, In Progress, Review, Done
  • Integrated MCP server with per-board endpoints, bearer token authentication, and public/private toggles

Additional capabilities include drag-and-drop task movement using vuedraggable, dual Kanban and Task List views, correction tasks with linked child tasks, configurable global and per-board MCP instructions, task deep-linking, and automatic real-time polling to synchronize human and AI actions.

Getting it running

The fastest path is Docker. With Docker installed, create a .env file from .env.example and set NUXT_SESSION_PASSWORD to at least 32 characters. Then start the stack:

docker-compose up -d

or

docker compose up -d

The application becomes available at http://localhost:3000. This stack includes the database, Adminer, and Mailpit.

For manual development, install dependencies and run the dev server. Typical steps are:

npm install
npm run dev

Environment variables must still be configured, and MySQL with Drizzle ORM must be available for persistence.

Who this is for

Moo Tasks suits developers and small teams that want lightweight, self-hosted task tracking with direct AI agent integration. It is practical for workflows where humans manage tasks visually while LLMs operate on them through the MCP protocol. Projects that need per-board MCP endpoints, controlled via bearer tokens and public/private flags, can use it as a coordination layer between human planning and automated execution. If you require email-based invitations and role management without a SaaS platform, it provides those primitives.

How it compares

Compared to general-purpose kanban tools, Moo Tasks adds an MCP server for programmatic AI access, which most alternatives lack. It is lighter than full project management suites but heavier than minimal MCP servers that expose no UI. Tools focused exclusively on LLM tool servers may offer less task-board polish, while traditional kanban applications typically omit bearer-token-gated MCP endpoints and correction-task workflows. The combination of Vue, Nuxt 4, Tailwind CSS v4, and MySQL via Drizzle ORM places it in a conventional web stack rather than a minimal container or serverless function.

The project is MIT-licensed and maintained at github.com/dizlexic/moo-tasks, with documentation at mootasks.dev.

View on GitHub →Official Website →