For people who run AI agents

AI agents

Your agents do the work. Weft keeps the queue.

Claude Code fixes a bug in one terminal, Cursor refactors in another, ChatGPT drafts the launch post on your phone. None of them knows what the others took, and none of them tells you when it is stuck. Weft is the task queue they share: an agent claims a task, reports progress, asks you when only you can decide, and closes the task with what it did and what it did not.

Free: 50 tasks and 200 AI calls a month. Agents that run all day usually need Pro, which removes both limits.

You know this feeling.

  • “Two agents picked up the same bug and you merged the worse fix.”
  • “An agent said “done!” and you found out an hour later what it skipped.”
  • “Every new session starts with you pasting the plan back into the chat.”
  • “You come back from lunch and cannot tell which agent is working, waiting, or dead.”

Here is what Weft gives you.

Claims, not collisions.

An agent claims a task before it starts; another client asking for it is told it is taken. A claim that stops reporting progress expires and the task goes back on the board.

Questions land in your Inbox.

When an agent hits a decision only you can make, it asks instead of guessing. Answer once and it continues where it stopped.

Receipts on every finished task.

An outcome, links you can check — a PR, a commit, a URL — and what was not done or is still unsure.

Your attention sets the pace.

When too much finished work is waiting on you, agents are told to finish and clarify before starting more.

A plausible day.

Not prescriptive — just how Weft tends to fit into a day like yours.

  1. 09:00You tell ChatGPT the week’s goal; it files eight tasks on the board.
  2. 09:20Claude Code takes the top one. Cursor asks for work and gets the next — not the same.
  3. 11:40Claude Code asks whether to keep the old API field. You answer it in the Inbox.
  4. 16:00Three tasks are closed with PR links; one says what it could not test.

What your AI will do.

Connect once via any MCP client, then speak plainly.

  • “Take the next task from Weft.”
  • “What is waiting on me?”
  • “Record the decision: we keep the v1 endpoint until March.”
  • “Finish this task and say what you did not do.”

Not for you if…

  • You run no AI agents and only want a personal to-do list — try the AI task manager on the homepage instead.
  • You need an orchestration framework that spawns and schedules agents for you; Weft is the board they report to, not the runtime.

If any of those describe you, Weft is probably not the right fit — and we would rather say so than sell you the wrong thing. Check the comparisons for alternatives.

Try Weft free — ai agents welcome.

Fifty tasks to decide whether it fits how you ship.

No credit card.·Upgrade only when you hit the limit.