Chrome Debug Mcp

Name: Chrome Debug Mcp
Author: raultov

@raultov

Chrome Debug Mcp について

chrome-debug-mcp is an asynchronous Rust-based Model Context Protocol (MCP) server that allows AI agents and Large Language Models to natively debug Chromium-based browsers via the Chrome DevTools Protocol (CDP).

#browser-automation #developer-tools #web-scraping

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

raultov

投稿者

Raúl Tovar

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-debug-mcp": {
      "command": "chrome-debug-mcp",
      "args": []
    },
    "chrome-docker": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "chrome-debug-mcp:v1.0.0",
        "--headless"
      ]
    },
    "chrome-docker-hybrid": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "--net=host",
        "chrome-debug-mcp:v1.0.0",
        "--host",
        "127.0.0.1"
      ]
    }
  }
}

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 Debug Mcp?

Chrome Debug Mcp is an asynchronous Rust-based Model Context Protocol (MCP) server that allows AI agents and large language models to natively control, automate, and debug Chromium-based browsers via the Chrome DevTools Protocol (CDP). Using cdp-lite, it hooks directly into the browser without heavy abstractions, and starting from v0.2.0 can manage the Chrome process lifecycle automatically.

How to use Chrome Debug Mcp?

Install via pre‑compiled binaries from the Releases page, via Cargo (cargo install --git https://github.com/raultov/chrome-debug-mcp), or with a shell script (Unix). Configure any MCP client (Claude Desktop, Gemini CLI, ChatGPT CLI) using a JSON configuration that specifies the executable or Docker command. You can run it natively, in Docker (headless), or in a hybrid mode where the container controls a Chrome instance on the host.

Key features of Chrome Debug Mcp

Local‑only navigation restriction (--local) for security
Send raw CDP commands (send_cdp_command) as a fallback
Automatic Chrome instance management (auto‑launch, restart, stop)
Proxy authentication handling (enable_proxy_auth)
User input simulation (click, fill, scroll)
Network log interception with advanced filtering
Console log retrieval with level filtering
Performance metrics and Core Web Vitals profiling
Live debugging with breakpoints, step‑over, and resume

Use cases of Chrome Debug Mcp

Debugging web applications from an LLM‑powered editor or chat interface
Automated browser testing with natural language instructions
Performance profiling and Core Web Vitals analysis
Real‑time inspection of network traffic and console errors
Controlled browser automation in Docker‑based CI/CD pipelines

FAQ from Chrome Debug Mcp

What are the runtime dependencies?

The MCP server requires the Rust stable toolchain (if building from source) and a Chromium‑based browser installed on the system. The server automatically finds Chrome in standard OS locations, or you can set the CHROME_PATH environment variable. For Docker usage, Chrome is included in the container.

How do I run Chrome in headless mode?

Use the --headless flag when starting the server: chrome-debug-mcp --headless. This is essential for Docker or server environments where no GUI is available.

Can I restrict navigation to local addresses only?

Yes, run with the --local argument. This restricts navigation to localhost, 127.0.0.1, 192.168.x.x, or .local addresses, which is recommended for securely debugging local development environments.

How do I connect to a Chrome instance on a different machine?

Use the --host argument to specify the target host (default 127.0.0.1). For example, --host host.docker.internal from inside a Docker container connects to a Chrome instance running on the host machine.

What is the `send_cdp_command` tool for?

It is an experimental tool that sends any raw CDP command directly to the browser. It serves as a powerful fallback for any domain or command not yet natively implemented in the specialized tools.