Core concepts

A run is one execution of a task. You give it a plain-language goal and an optional output shape; BrowserPilot plans the steps, drives a managed browser, and returns a result. A run has a status — queued, running, completed, failed, or cancelled — and is the unit you create, observe, and bill against.

Every run executes inside an isolated browser session. The session is created fresh, streams live while the run works, and is torn down afterward — so jobs never collide or leak state. The session's frames are retained for replay.

A run is a sequence of steps — navigate, click, type, extract. Each step records the action, a screenshot, the agent's reasoning, and its token and action cost. Steps are why a failed run is a recording you can scrub through rather than a black box.

A webhook is an endpoint you register to receive run events. Events are HMAC-signed, retried with backoff, and fanned out to every endpoint in your workspace. See the Webhooks guide.

Cost accrues per step and rolls up per run, project, and plan. Each plan carries a monthly quota of runs; quota usage is visible live in the dashboard, and a run that would exceed it is rejected rather than silently billed.

A template is a saved run configuration you can re-run on demand or on a schedule. A share link is a public, read-only replay of a single run — revocable, with no workspace access. Both are covered in their product pages.