This community node package adds a single Browser Use node to n8n, giving a workflow access to Browser Use Cloud, where AI agents drive managed browsers for web research, data extraction, form filling, testing, and multi-step tasks. The node can also be attached to an AI Agent node as a tool, so an agent can hand it a task in natural language.
One node covers three API generations, selected from an API Version dropdown; the node rewrites the trailing /api/vN segment internally, so an existing credential keeps working. Authentication is an API key sent in the X-Browser-Use-API-Key header.
| Version | Use it for | Notes |
|---|---|---|
| v4 (default) | New integrations, hard or long multi-step workflows | Highest accuracy. Not available on Zero Data Retention projects. |
| v3 | Cost- and speed-sensitive work | Session-based agents plus standalone cloud browsers. |
| v2 | Existing workflows only | Legacy; no longer actively maintained upstream. |
In v4 the unit of work is a run, and every run belongs to a session that keeps conversation context and browser state, so follow-up messages queued onto the session reuse both. Run operations cover dispatch-and-wait, create, get, get status, list, cancel, event streams, and attachments; session operations queue or cancel messages, optionally interrupting the active run, and purge sessions on Zero Data Retention projects. Run options include the model, provider-native model parameters, a session or workspace to continue, an LLM judge, a maximum cost in US dollars, and browser settings such as profile, proxy country, screen size, and recording.
Structured output differs between versions. API v4 has no server-side output schema, so when structured extraction is enabled the node appends the JSON Schema to the task as an instruction and parses the returned text into parsedResult, leaving the raw text in result and reporting a mismatch in structuredOutputError without failing the run. Use v3 when the API must enforce the schema itself.
Features
- Single node, three API versions: v4 runs, v3 sessions and browsers, and legacy v2 tasks selected from a dropdown
- Run and wait: dispatch a task and poll until it completes, fails, or is cancelled, then return the full run summary
- Session continuity: pass a session ID to reuse conversation context and browser state across runs
- Message queueing: queue follow-up instructions on a session, with an interrupt option that cancels the active run
- Model selection: GPT-5.5 and 5.6 family, Claude Opus and Sonnet, Gemini 3 variants, GLM, Grok, Kimi, and MiniMax on v4; Claude, Gemini Flash, and Browser Use models on v3
- Cost controls: a per-run maximum spend, plus token counts and total cost returned in the run summary
- Structured extraction: schema-enforced output on v3, best-effort JSON Schema parsing into parsedResult on v4
- Standalone browsers: provision a cloud browser and get liveUrl and cdpUrl for use from other automation tools
- Browser settings: profile reuse, proxy country or custom proxy, screen size, timeouts, and session recording
- Agent tool mode: attach the node to an AI Agent so an agent can invoke it directly
