Research / Browser Agent

Run long autonomous research and browser work without managing persistent workers.

How it works

1.A user requests something like "Research these 50 companies."

2.Your backend creates a Badgr Agent Job for the request.

3.Badgr runs the research/browser workload inside the cost and runtime limits you set.

4.Badgr returns whatever outputs the job produced — structured result, files, artifacts, logs, verification evidence, cost, and status.

5.Your application displays the result, and can submit a follow-up job (with parent_job_id) to continue or rerun the work.

Badgr owns

Isolated execution environment
Browser-capable agent runtime
Cost cap, runtime cap, cancel
Logs, status, artifact capture
Teardown and billing settlement

You own

What to research and how to phrase it
Which agent and budget to use
Displaying and acting on the result
Follow-up / continuation requests
Your product UI

Browser/research work runs as an Agent Job (agent: "playwright") — there is no separate browser job type. Only what's actually supported today is documented here.

Build this with Badgr — Agent Jobs quickstart →