name: googler description: "Scrape Google search results using MCP. Uses progressive scraping with patchright to bypass Google's bot detection. Use when: user asks to search Google, find information online, or get web search results. NOT for: API-based search (use Google Custom Search API), authenticated Google services, or high-volume automated queries." homepage: https://github.com/kakkoii1337/gai-mcp-googler metadata: { "openclaw": { "emoji": "🔍", "requires": { "node": ">=18.0.0", "npm": true }, "install": [ { "id": "npm", "kind": "npm", "package": "gai-mcp-googler", "label": "Install via npm", }, ], }, }

Googler Skill

Scrape Google search results via MCP (Model Context Protocol).

When to Use

✅ USE this skill when:

• "Search Google for..."
• "Find information about..."
• "What does Google say about..."
• "Search the web for..."
• Need web search results in Claude Code

When NOT to Use

❌ DON'T use this skill when:

• Need API-based search → use Google Custom Search API
• Authenticated Google services → use appropriate Google APIs
• High-volume automated queries → use dedicated search APIs
• Need cached/archived results → use Wayback Machine

Installation

npm install -g gai-mcp-googler

Or run directly:

npx gai-mcp-googler

MCP Configuration

Add to Claude Code config (~/.claude.json):

{
    "mcpServers": {
        "googler": {
            "type": "stdio",
            "command": "npx",
            "args": ["gai-mcp-googler"],
            "env": {}
        }
    }
}

How It Works

Progressive scraping to bypass Google's bot detection:

1. HTTP request - Fastest, but Google usually blocks it
2. Headless browser - Uses patchright (Chromium without detection)
3. Non-headless browser - Shows browser window, most reliable

Tools

ping

Health check endpoint.

Returns: "pong"

google_search

Scrape Google search results.

Parameters:

• query (required): Search query string
• num_results (optional): Number of results (1-10, default: 10)
• headless (optional): Use headless mode (default: true)

Example:

{
    "query": "current singapore time",
    "num_results": 5,
    "headless": false
}

get_lucky

Get first result directly (I'm Feeling Lucky).

Parameters:

• query (required): Search query string
• headless (optional): Use headless mode (default: false)

Example:

{
    "query": "wikipedia",
    "headless": false
}

Notes

• Non-headless mode is more reliable but shows browser window
• Headless mode may be blocked by Google
• Server logs written to /tmp/gai-mcp-googler.log
• Uses patchright (undetected chromedriver equivalent for Node.js)