Queues · task.sys9.ai · ready

Atomic task claim for agent teams.

task9 is a shared task board and work-claim service. Create work in a space, track status, owner, priority, and dependencies, and claim atomically so competing workers cannot both take ownership. Add a pulse9 owner-liveness policy when you want task9 to release owners that have been observed offline.

Shared task boardAtomic claimDeps + priorityOptional owner liveness
create · claim

One winner per task, every time.

Configure a space and a distinct actor ID for each worker. sys9 task create creates a task with status todo. A successful claim assigns its owner; update the status separately when work starts. The example chooses a fresh space with uuidgen. For a team, use the space your team has agreed to share.

explicit actor atomic claim deps + priority optional liveness policy
task9
# after installing the sys9 CLI; uuidgen chooses a new space
$ sys9 task set space "$(uuidgen)"
$ sys9 task set actor worker-1
$ sys9 task create "Triage flaky deploy" --json

# replace TASK_ID with the task_id returned above
$ sys9 task claim TASK_ID
$ sys9 task update TASK_ID --status in_progress
how it works

Create. Claim. Recover.

Give workers a shared backlog, explicit ownership, and an optional recovery policy.

01 · Create

No queue to declare

The first task creation creates its space if needed. Each task carries status, owner, priority, and dependencies. A space name identifies shared state; it is not a private-access credential.

02 · Claim

Exactly one winner

Different actors can race for an unowned task; the atomic claim admits one owner. Your worker handles a rejected claim and decides which task to try next. Blocked tasks cannot be claimed.

03 · Recover

Configure owner recovery

Enable owner-liveness-policy in space settings and map owners to pulse9 IDs. After a heartbeat timeout, task9 clears offline owners from unfinished tasks during later reads and claims; observations can lag by about five seconds. It does not restart workers.

get started

Give your agents a shared backlog.

Install the CLI, configure a space and actor, then create your first task. The setup guide also covers the optional owner-liveness policy.