Create Browser Use Run
Start a browser-use run, or continue a disposable session (session_id).
Async: accepts the task and returns a slim RunAccepted handle (202). Poll
GET /runs//status or GET /runs/ for progress + result.
MOCKED dispatch. A stray workspaceId is rejected by the request model’s
validator (shared agent-workspaces are browsercode-only); files still work via
attachedFileIds.
Authorizations
Body
Start a new browser-use run OR continue a disposable session.
SIMPLER than browsercode: NO workspace concept (shared agent-workspaces are
browsercode-only) and NO code. Sending workspaceId is rejected (see the
validator) — the capability model is enforced, not silently ignored.
Continuation via session_id (v2 persist-memory). browser-use 409s on a busy
session (it does not queue — that's browsercode's behavior).
What you want the agent to do in the browser.
1Browser/proxy/screen settings for this run.
Continue a disposable session. 409 if that session is currently busy (browser-use does not queue). Omit to start fresh.
Model to run. Omit for the server default.
Upload ids from POST /agents/browser-use/files to give this run.
20Response
Successful Response
Async run accepted. Poll GET /runs/{id}/status or GET /runs/{id} for progress + result.
Run id, e.g. "run_...".
Which agent is running it.
browser-use, browsercode The conversation this run belongs to (new or continued).
Initial state — 'queued' or 'running'.
queued, running, completed, failed, stopped The run's workspace, if the agent has one (browsercode). Its files live here.
Attached upload ids that could not be confirmed in storage at creation. The run still starts.