A to-do list built for a world with AI teammates
Most task managers assume every task gets done by the person who wrote it down. Done doesn't.
We kept ending up with two separate lists: the tasks we tracked for ourselves, and the ad-hoc prompts we'd fire off to an AI agent and then lose track of. Done merges them. Every task lives in one place, with one owner, and an assignee that can be you or Claude.
When a task is assigned to Claude, an agent connects to your account over its own MCP endpoint, picks up ready work, and reports back into that task's activity feed — comments, status changes, a final summary. You stay the owner. Assigning to Claude is delegation, not a hand-off you lose visibility into.
Done is deliberately small: no OKRs, no Gantt charts, no notifications you have to tune out. Quick capture, projects, recurring and dependent tasks, and a daily digest — the parts of a task manager that earn their keep every day.
How it's built
Done runs on a static Next.js front-end backed by a single lightweight API service that serves the REST API and a remote MCP server, both backed by one database. No separate backend to operate, no cold starts to wait out.
It's a small project, actively evolving. If you run into something rough, or want to tell us what's missing, reach out — we read everything.
Who's behind Done
Done is built by Alejandro Rioja, founder of Done. You can find more of his work and get in touch at alejandrorioja.com.