generated: '2026-08-29'
method: probed
source: >-
  POST {"jsonrpc":"2.0","id":1,"method":"tools/list"} to https://api.browser-use.com/mcp and
  https://api.browser-use.com/v3/mcp; docs at https://docs.browser-use.com/cloud/guides/mcp-server
status: live
deployment:
  mode: remote
  endpoint: https://api.browser-use.com/mcp
  auth: oauth
  verified: probed
  note: >-
    tools/list answers 200 anonymously, which is how the tool schemas below were captured verbatim.
    Tool CALLS are protected: /.well-known/oauth-protected-resource names
    https://api.browser-use.com/mcp as the resource, https://api.browser-use.com as its authorization
    server, and mcp as the required scope. The documented client configurations instead pass an
    x-browser-use-api-key header, so both an OAuth bearer and a project API key are accepted. There is
    no first-party stdio package — nothing here needs npx or a local install.
servers:
- id: browser-use-mcp
  name: Browser Use MCP
  endpoint: https://api.browser-use.com/mcp
  transport: http
  http_status: 200
  server_info:
    name: browser-use-web-automation
    version: 1.0.0
    description: >-
      AI-powered web browser automation for scraping, data extraction, form filling, and general web
      tasks
  protocol_version: 2024-11-05
  auth:
    oauth:
      authorization_server: https://api.browser-use.com
      scope: mcp
      metadata: well-known/browser-use-oauth-authorization-server.json
    api_key:
      header: x-browser-use-api-key
  tools_file: mcp/browser-use-mcp-tools.json
  tool_count: 6
  tools:
  - name: browser_task
    summary: >-
      Create a browser automation task; spins up a full cloud browser and executes steps. The tool
      description instructs the client to check list_skills first.
    parameters: ["task", "max_steps", "profile_id", "model", "proxy_country"]
    required: ["task"]
  - name: monitor_task
    summary: >-
      Get current status and progress of a browser automation task, including the most recent step
      with full agent reasoning (evaluation, memory, actions).
    parameters: ["task_id"]
    required: ["task_id"]
  - name: list_browser_profiles
    summary: >-
      List all available cloud browser profiles for the authenticated project.
    parameters: []
    required: []
  - name: list_skills
    summary: >-
      Find a pre-built skill that already performs the requested website action before falling back to
      a full browser_task.
    parameters: []
    required: []
  - name: get_cookies
    summary: >-
      Extract cookies from a browser profile for authenticated skill execution.
    parameters: ["profile_id", "domains"]
    required: ["profile_id"]
  - name: execute_skill
    summary: >-
      Execute a skill with provided parameters; skills run without browser overhead.
    parameters: ["skill_id", "parameters", "session_id"]
    required: ["skill_id"]
- id: browser-use-mcp-v3
  name: Browser Use MCP (v3 session API)
  endpoint: https://api.browser-use.com/v3/mcp
  transport: http
  http_status: 200
  documented_at: https://docs.browser-use.com/cloud/guides/mcp-server
  auth:
    api_key:
      header: x-browser-use-api-key
  tools_file: mcp/browser-use-v3-mcp-tools.json
  tool_count: 7
  tools:
  - name: run_session
    summary: Create a new browser agent session and run a task.
    parameters: ["task", "model", "profile_id", "workspace_id", "keep_alive", "output_schema", "proxy_country"]
    required: ["task"]
  - name: send_task
    summary: Send a follow-up task to an existing keep-alive session.
    parameters: ["session_id", "task", "keep_alive", "output_schema"]
    required: ["session_id", "task"]
  - name: get_session
    summary: Get current status and output of a session.
    parameters: ["session_id"]
    required: ["session_id"]
  - name: get_session_messages
    summary: Get the agent messages/steps for a session.
    parameters: ["session_id", "after", "limit"]
    required: ["session_id"]
  - name: stop_session
    summary: >-
      Stop a running session. strategy "task" stops only the task; "session" destroys the sandbox.
    parameters: ["session_id", "strategy"]
    required: ["session_id"]
  - name: list_sessions
    summary: List recent sessions for the authenticated project, with status and cost info.
    parameters: ["page", "page_size"]
    required: []
  - name: list_browser_profiles
    summary: List all available cloud browser profiles for the authenticated project.
    parameters: []
    required: []
- id: browser-use-docs-mcp
  name: Browser Use Documentation MCP (Mintlify)
  endpoint: https://docs.browser-use.com/mcp
  transport: http-sse
  http_status: 200
  auth:
    none: true
  tool_count: 2
  tools:
  - name: search_browser_use
    summary: Search the Browser Use knowledge base for docs, code examples, API references and guides.
  - name: fetch_browser_use
    summary: Fetch a full Browser Use documentation page by URL.
  note: >-
    This is the Mintlify docs server, not a product surface. Recorded because it is a real anonymous
    MCP endpoint an agent can call, but it operates on documentation, not on the Browser Use account.
client_configuration:
  documented: https://docs.browser-use.com/cloud/guides/mcp-server
  claude_code: >-
    claude mcp add -t http -H "x-browser-use-api-key: YOUR_API_KEY" browser-use
    https://api.browser-use.com/v3/mcp
  note: >-
    The published docs page still configures the /v3/mcp endpoint. The /mcp endpoint carries the newer
    skills-aware tool set and is the resource named by the OAuth protected-resource metadata.
gaps:
- >-
  Two live product MCP endpoints with different tool sets and no published deprecation note on either.
  A client following the docs gets the v3 session tools; a client following the well-known OAuth
  metadata gets the skills tools.