Browser Use

Browser Use for developers

Two ways to build with Browser Use.

Use our hosted web agents end-to-end, or run your own code on our production browser infrastructure.

01Fully managed

Hosted Web Agents

Give Browser Use a task. Get completed work.

POST /api/v4/runs complete
task"Research this company"resultstructured output
Run an agent Explore hosted agents
API V4PythonTypeScript
02Bring your own agent

Browser Infrastructure

Production browsers for your own code: Playwright, Puppeteer or any CDP client, live view, recordings.

browser = browsers.create() connected
stealthproxyprofile
Launch a browser Explore infrastructure
SDKRESTCDP

Browser Infrastructure

Script what you can. Hand the rest to an agent. Watch all of it.

  1. 01

    Deterministic steps

    Create a browser and attach Playwright or Puppeteer to its cdpUrl. Your selectors and waits run unchanged on a real Chromium.

    Playwright and Puppeteer
  2. 02

    AI steps, same browser

    Hand the same cdpUrl to a browser-use agent with BrowserSession(cdp_url=...) for the pages you cannot script, then take control back.

    Example
  3. 03

    Live view and replay

    Every browser returns a liveUrl to watch or embed. Create it with enableRecording to get an mp4 after it stops.

    Live view and recording

Interfaces

Developer toolkit

Pick the interface your stack already speaks.

API V4DocsOpenAPI 3.1JSONCloud SDKGitHubWebhooksGuideMCP serverMCPAgent-readable docsTXT

Authentication

API keys for REST. OAuth for MCP.

REST + SDK

API key

Send X-Browser-Use-API-Key with API V4 requests.

Create a key

MCP clients

OAuth 2.0

Authorization Code with PKCE. The published permission is the mcp scope.

OAuth metadata