Daniel Roe, a lead maintainer of the Nuxt framework, shipped something worth paying attention to. Moo Tasks is a kanban board with a twist: it has a built-in MCP server that lets AI agents connect, pick up tasks, and update them alongside human teammates. You drag cards around in the browser. Your AI assistant does the same thing through an API. The Model Context Protocol is still early, but projects like this show where it's heading. MCP gives AI agents a standard way to interact with external tools and data. Skrun extends this by deploying agent skills as an API, creating callable endpoints for your workflows. Moo Tasks bets that agents need to operate in the same project management flow as people, not in a separate chat window. Waffle auto-tiles your AI agent terminals with zero config, offering a visual workspace for managing parallel processes. Technically, Roe built this on fresh tooling: Nuxt 4, Tailwind CSS v4, MySQL with Drizzle ORM. The stack runs in Docker or standalone with Node.js 22+. It's MIT licensed. The five-column workflow (Backlog, To Do, In Progress, Review, Done) won't surprise anyone who's used JIRA or Trello. But the MCP integration is what sets it apart. You can configure per-board instructions that tell agents how to behave. A task-workflow prompt guides agents through discovering and completing work. The question behind Moo Tasks is whether shared human-agent task boards become a standard pattern. Right now, most AI coding assistants work in isolation. You chat with them, they do something, and there's no persistent coordination layer.