Skip to directory

Browser Use MCP

Browser Use

stdio (local); Streamable HTTP (cloud)21+ documented tools16

About Browser Use MCP

Browser Use exposes browser automation to AI clients through either a local Python MCP server or a hosted cloud MCP service, with tools for browser actions and autonomous tasks.

What you can use it for

  • Automate browser navigation, clicking, typing, scrolling, tab management, and content extraction
  • Delegate multi-step browser tasks to an agent
  • Use hosted profiles/sessions and monitor cloud task progress
Connection and setup
Hosting
Local or remote cloud-hosted
Transport
stdio (local); Streamable HTTP (cloud)
Endpoint
https://api.browser-use.com/mcp
Authentication
Cloud endpoint uses x-browser-use-api-key and incurs documented per-task/per-step charges. Local server requires an LLM API key (OPENAI_API_KEY or ANTHROPIC_API_KEY); browser-use cloud key is needed only for cloud browser usage.

Requirements

  • Python/uvx for local server
  • LLM API key for local agent
  • Chrome/browser setup as required by chosen execution mode
  • Browser Use API key and paid usage for cloud MCP

Setup steps

  1. Choose local or cloud mode deliberately; local server command is uvx --from "browser-use[cli]" browser-use --mcp.
  2. For local mode, use a documented LLM provider key via a secret environment variable and configure access safeguards.
  3. For cloud mode, add https://api.browser-use.com/mcp and x-browser-use-api-key header using a secret; review task and step charges before use.
  4. Restart/reload the MCP client and verify a safe read-only page navigation.
  5. For logged-in browsing, confirm the browser/profile context and allowed-domain controls before asking for actions.
Tools 21+

Stored tools combine several Browser Use MCP surfaces and releases. Current local MCP-native and CLI 3.0 surfaces differ; consult the selected installation’s tools/list.

  • browser_task
    01
  • execute_skill
    02
  • list_skills
    03
  • get_cookies
    04
  • list_browser_profiles
    05
  • monitor_task
    06
  • retry_with_browser_use_agent
    07
  • browser_navigate
    08
  • browser_click
    09
  • browser_type
    10
  • browser_get_state
    11
  • browser_scroll
    12
  • browser_go_back
    13
  • browser_list_tabs
    14
  • browser_switch_tab
    15
  • browser_close_tab
    16
  • browser_extract_content
    17
  • browser_screenshot
    18
  • browser_list_sessions
    19
  • browser_close_session
    20
  • browser_close_all
    21

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

Limitations and notes
  • Cloud browser tasks are chargeable: docs list browser_task at $0.01 plus per-step and execute_skill at $0.02; verify current account pricing before use.
  • Cloud requires an API key; local mode relies on an LLM provider key and may not share the cloud session/profile.
  • A browser agent can submit forms or otherwise change external state; require explicit review before consequential actions.
  • Tool sets differ between hosted cloud MCP and local stdio mode.
Original sources
  • official tool reference

    The current repository has multiple browser-use MCP surfaces (local MCP-native agent and CLI 3.0). Stored rows represent mixed surfaces, so no universal total is asserted.

  • official documentation

    The current repository has multiple browser-use MCP surfaces (local MCP-native agent and CLI 3.0). Stored rows represent mixed surfaces, so no universal total is asserted.

Documentation checked 2026-10-28T13:02:00.000Z. Connection has not been runtime tested.

Similar tools