Browser Use MCP Server

@MCP-Mirror

リポジトリ

基本情報

カテゴリ

ブラウザ自動化

ライセンス

MIT license

トランスポート

stdio

作者

MCP-Mirror

Browser Use MCP Server について

Mirror of

#browser-automation #ai-agents#web-control

Browser Use MCP Server の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "co-browser_browser-use-mcp-server": {
      "command": "uv",
      "args": [
        "tool",
        "install",
        "mcp-proxy"
      ]
    }
  }
}

Browser Use MCP Server のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is browser-use-mcp-server?

An MCP server that enables AI agents to control web browsers using the browser-use library. It provides a bridge between AI assistants and browser automation, allowing agents to navigate, extract data, and perform actions on web pages.

How to use browser-use-mcp-server?

Install prerequisites: uv, playwright, and mcp-proxy. Set up a .env file with OPENAI_API_KEY (and optionally CHROME_PATH and PATIENT). Install dependencies with uv sync, pip install playwright, and playwright install. Run in SSE mode with uv run server --port 8000 or in stdio mode by building the package and running browser-use-mcp-server run server --port 8000 --stdio --proxy-port 9000. Docker is also supported.

Key features of browser-use-mcp-server

Browser automation controlled by AI agents
Dual transport: SSE and stdio protocols
Real-time VNC streaming to watch automation
Async task execution for browser operations

Use cases of browser-use-mcp-server

AI assistant navigating a website and returning top-ranked content
Automating form filling, searches, or data extraction via natural language
Debugging or monitoring web automation in real-time through VNC

よくある質問

What prerequisites are needed?

You need `uv` (fast Python package manager), `playwright` (browser automation), and `mcp-proxy` (for stdio mode). Install them with the provided commands.

How do I configure the server for different MCP clients?

Configuration paths vary by client: for Cursor use `./.cursor/mcp.json`, for Windsurf `~/.codeium/windsurf/mcp_config.json`, and for Claude Desktop the path depends on OS (Mac or Windows).

How can I run browser-use-mcp-server with Docker?

Build the image with `docker build -t browser-use-mcp-server .` and run with `docker run --rm -p8000:8000 -p5900:5900 browser-use-mcp-server`. You can optionally set a custom VNC password via a mounted secret file.

What is the default VNC password and how do I watch the browser?

The default VNC password is `browser-use`. To watch, use a VNC viewer such as noVNC by cloning it and pointing to `localhost:5900`.

What transport modes are supported?

Both SSE (Server-Sent Events) and stdio modes are supported. SSE runs directly from source; stdio requires building and installing the package as a global tool.