Stealth Browser Workflows
Typical workflow
- Create a session with
POST /v1/sessions. - Wait until
stateisready(or poll if you usedwait=false). - Connect with
connect_url(Chromium CDP) or the Playwright websocket for Firefox. - Optionally open
live_view_urlto watch or take control. - End the session with
DELETE /v1/sessions/{session_id}when you are done.
What you get
| Piece | Role |
|---|---|
| REST API | Create, list, get, and delete sessions; profiles; usage; countries |
connect_url / ws_url | Drive the browser from code |
live_url / live_view_url | Watch the screen; take control in the view page |
| MCP | Staging host + /mcp with X-API-Key (prod MCP after activation) |
Stealth Browser is not the Scraper API. It uses its own host and plan. For now we check staging only; production comes after activation. Extract, Search, and Light stay on scraper.geonode.io.
Engines
| Engine | Notes |
|---|---|
chromium (default) | CDP endpoint. Profiles, resource blocking, Stagehand, and page counts apply. |
firefox | Playwright server (not CDP). Uses two plan slots. No profiles, resource blocking, or Stagehand. Page count is null. |
Plan limits to watch
Read plan on the session response (and usage) rather than hard-coding limits:
- Session slots (
max_sessions/slots_*). Firefox uses two slots. - Max TTL (
max_ttl_seconds) - Per-session traffic (
max_session_mb, megabytes of 1,000,000 bytes) - Monthly traffic (
monthly_gb) when the plan sets one
Was this page helpful?