Google Search MCP Server

A powerful Model Context Protocol (MCP) server providing Google search and image search capabilities for Claude Code

Features

🌐 Google Text Search

• Search Google with customizable result count
• Extract titles, links, and descriptions
• Support for multiple languages

📄 Web Content Extraction

• Extract main content from any webpage
• Smart content filtering (ads, navigation removed)
• Batch extraction support

🔍 Combined Search & Extract

• Search and auto-extract multiple results
• One-click deep content retrieval

🖼️ Image Search with AI Analysis ⭐

• Dual-platform search: Google Images + Pinterest
• Multi-model AI visual analysis: Supports GLM-4.6V, GPT-4O, Claude 3.5, and more
• Quality scoring system: S/A/B/C grading
• Perfect for finding product reference images

Installation

Prerequisites

• Node.js >= 18.0.0
• Claude Code with MCP support

Setup

1. Clone the repository

git clone https://github.com/hinsmoment/google-search-mcp.git
cd google-search-mcp

2. Install dependencies

npm install

3. Configure Claude Code

Add to your Claude configuration (~/.claude.json or via Settings):

{
  "mcpServers": {
    "google-search": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/google-search-mcp/server.js"]
    }
  }
}

4. Restart Claude Code

Usage

1. Google Search

Search Google for "Claude Code AI assistant"

2. Extract Webpage Content

Extract content from https://example.com

3. Batch Extract URLs

Extract content from these URLs:
- https://github.com/topic1
- https://github.com/topic2
- https://github.com/topic3

4. Image Search (Product Reference)

Find similar products using this image: /path/to/product.jpg

MCP Tools

| Tool | Description | Parameters | |------|-------------|------------| | google_search | Google text search | query, maxResults (1-25) | | fetch_url | Extract webpage content | url | | batch_fetch_urls | Batch extract multiple URLs | urls (array) | | search_and_extract | Search + extract results | query, maxResults (1-10) | | search_by_image | Image-based product search | image_path, max_results, enable_ai_analysis |

Image Search Details

How It Works

┌─────────────────┐
│  Upload Image   │
└────────┬────────┘
         ↓
┌─────────────────┐
│  AI Analysis    │
│  (Multi-Model)   │
│  - Product type │
│  - Keywords     │
└────────┬────────┘
         ↓
    ┌────┴────┐
    ↓         ↓
┌───────┐  ┌───────┐
│Google │  │Pinterest│
│Images │  │Search  │
└───┬───┘  └───┬───┘
    └────┬────┘
         ↓
┌─────────────────┐
│  Merge & Score  │
│  S/A/B/C Grades │
└─────────────────┘

Quality Scoring

| Grade | Score | Sources | |-------|-------|---------| | S | 85-100 | Tmall HK, Tmall, JD Mall, Brand Sites | | A | 75-84 | JD, Amazon, Taobao, Pinterest | | B | 60-74 | Other e-commerce sites | | C | < 60 | General websites |

Technical Architecture

Browser Automation

• Puppeteer (primary)
• Playwright (automatic fallback)

Key Dependencies

{
  "puppeteer": "^22.15.0"
}

Module Structure

google-search-mcp/
├── server.js           # MCP server implementation
├── lib/
│   └── image-search.js # Image search module with AI analysis
├── package.json
└── README.md

Configuration

Optional: AI Analysis API Key

For enhanced image analysis, set your GLM API key:

# Linux/macOS
export GLM_API_KEY=your_api_key_here

# Windows (PowerShell)
$env:GLM_API_KEY="your_api_key_here"

# Windows (CMD)
set GLM_API_KEY=your_api_key_here

Note: The fallback keyword-based search works without an API key.

Examples

Example 1: E-commerce Product Research

Find similar products to: ./images/dinosaur-pencil-case.jpg
Return top 15 results with AI analysis enabled

Output includes:

• Product image analysis
• Keywords extracted
• Dual-platform results (Google + Pinterest)
• Quality scores with grades

Example 2: Content Research

Search for "MCP protocol Model Context"
Extract full content from top 5 results

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Development Setup

# Install dependencies
npm install

# Run tests
npm test

# Start MCP server (for debugging)
node server.js

Coding Standards

• Follow existing code style
• Add comments for complex logic
• Test your changes thoroughly

Troubleshooting

Puppeteer Chrome Download Issues

If Puppeteer fails to download Chrome:

# Skip Chrome download and use system Chrome
PUPPETEER_SKIP_CHROMIUM_DOWNLOAD=true npm install puppeteer

MCP Server Not Starting

1. Check Node.js version: node --version (requires >= 18.0.0)
2. Verify file path in Claude configuration
3. Check Claude Code logs for errors

Image Search Returns No Results

• Verify image path is correct
• Check internet connection
• Try disabling AI analysis if API key issues

Limitations

• Google search may be affected by anti-scraping measures
• Pinterest direct scraping is limited (uses Google site:pinterest.com filter)
• API-based AI analysis requires separate GLM API key

Roadmap

• [ ] Add more image search platforms (Behance, Dribbble)
• [ ] Implement local image similarity matching (CLIP/DINOv2)
• [ ] Add result caching
• [ ] Support for more AI vision models

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Model Context Protocol by Anthropic
• Puppeteer for browser automation
• Zhipu AI for GLM-4.6V vision model
• OpenAI for GPT-5.2 / GPT-4O vision model
• Anthropic for Claude 4.5 Sonnet vision model

Author

Created by 成哥

Special thanks: Developed with Claude Code assistance

Support

• 🐛 Report bugs: GitHub Issues
• 💡 Feature requests: GitHub Discussions