MCP.so
ログイン

Chrome MCP Server

@jverre

Chrome MCP Server について

MCP server to interact with Chrome

基本情報

カテゴリ

その他

ライセンス

MIT

ランタイム

python

トランスポート

stdio

公開者

jverre

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": {
    "chrome-mcp-server": {
      "command": "uv",
      "args": [
        "run",
        "uvicorn",
        "demo_implementation.main:app",
        "--reload"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

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 Chrome MCP Server?

Chrome MCP Server is an implementation of the Model Context Protocol (MCP) designed to interact with Chrome. It was built as a learning exercise to understand the MCP protocol and to provide a server that can be used in Cursor to take screenshots of a page and validate implementation.

How to use Chrome MCP Server?

For the demo implementation, run uv run uvicorn demo_implementation.main:app --reload and connect via the MCP inspector at http://0.0.0.0:8000. For the robust implementation (which uses the MCP Python SDK), start the server with mcp run app/main.py --transport sse and then use the MCP inspector.

Key features of Chrome MCP Server

  • Provides an MCP implementation independent of Anthropic’s library.
  • Supports taking screenshots of a webpage.
  • Allows validation of implementation in Cursor.
  • Includes a demo implementation with initialization over SSE and POST endpoints.
  • Offers a more robust implementation using the MCP Python SDK.

Use cases of Chrome MCP Server

  • Taking screenshots of pages for visual verification.
  • Validating implementation in Cursor during development.
  • Learning the MCP protocol through a hands‑on example.

FAQ from Chrome MCP Server

Is Chrome MCP Server a reference implementation of the MCP protocol?

No, it is not a reference implementation. It was built as a learning exercise to understand the protocol.

What are the limitations of the demo implementation?

The demo implementation has several limitations, including lack of error handling and inadequate cleanup of sessions.

How does the initialization process work in the demo implementation?

Initialization is done in three steps: 1) an initial call to /sse to get the session URI, 2) the client sends an initialize message, and 3) the client sends a notifications/initialized message.

What transport does the robust implementation use?

The robust implementation uses SSE (Server‑Sent Events) transport.

コメント

「その他」の他のコンテンツ