A task queue for your AI coding agents

Write down the next useful request when you understand it. The queue keeps that request attached to the correct repository until its provider capacity is ready.

Make the request reviewable

Choose a project and provider, describe the intended behavior, and include relevant references or checks. A focused request is easier to inspect after execution than a long list of unrelated changes. Star important tasks or rename them to keep their purpose clear.

Adjust waiting work without interrupting the current run

Edit or reorder queued prompts as priorities change. Run now gives urgent work dispatch priority, subject to required capacity. It is not a way to bypass the provider limits or make a conflicting edit safe.

Preserve a record after the process ends

The task retains its prompt, activity, result, and supported conversation information. Search can find work by names, requests, follow-ups, responses, results, and errors. Use the record to decide whether to continue the same thread or start fresh.

Know what the queue does not decide

The queue schedules capacity. You still decide dependencies, repository ownership, and when an implementation is ready for release. A clear request and deliberate review make the scheduling useful. Without them, more dispatched prompts can simply create more unfinished decisions.

Go deeper

Languages