Sixpent AI Jobs

Ready.

Projects

Name System key

Sites

Every site this project can be answered on, with its own URL. All are ticked to start with: more sites is more rate-limit budget, and a conversation moves to another one — carrying its history — when the site it is on starts refusing. Untick a site to stop using it, or leave its URL empty and it is left out.

Rotate between sites (aggressive) Off, jobs stay on the first working site and a conversation stays where it is — one site absorbs the project until it starts refusing. On, every job goes to whichever site this project used least recently, so five rate-limit budgets are spent evenly instead of one to exhaustion. The cost: a conversation moves each turn and replays its recent history into the new thread, which is slower and reads back a little differently.

Enabled

Fleet settings

Every site rate limits the browser, not this API, and each one has its own budget. These slow the workers down; they take effect on each worker's next heartbeat, within 30 seconds. Thinking defaults are frozen onto each new job as soon as it is created.

Seconds between jobs Gap before one browser starts its next job. 0 disables pacing. Cooldown after a rate limit Seconds a worker stops claiming for that site after it is limited. Max concurrent jobs per worker Fleet-wide ceiling. 0 lets each worker use its own setting.

Per-site overrides

Blank inherits the fleet default above. ChatGPT objects to back-to-back page loads long before Claude does, so one set of numbers for all five is either too slow somewhere or too fast somewhere else.

Provider thinking defaults

Used when jobs-create omits a provider override. Existing queued jobs keep the values they were created with. Claude also enables Thinking.

Jobs

Status Project

Sixpent AI Jobs v…

Projects

Fleet settings

Jobs

Extension Clients

Job details

Admin Login

Projects

Fleet settings

Jobs

Extension Clients