MCP Camoufox Scraper Server

A proof-of-concept MCP (Model Context Protocol) server that uses Camoufox for web scraping with JavaScript disabled and network request monitoring.

🚀 Quick Start

Prerequisites

• Python 3.10 or higher
• Poetry (Python dependency manager)
• macOS, Linux, or Windows

1. Clone/Download the Project

git clone <your-repo> mcp-camoufox-scraper
cd mcp-camoufox-scraper

2. Install Poetry (if not already installed)

curl -sSL https://install.python-poetry.org | python3 -

3. Install Dependencies

poetry install

4. Verify Setup

poetry run python setup_verify.py

You should see all checks pass:

🎉 Setup verification successful!

5. Run Full Test

poetry run python test_mcp_server.py

You should see:

🎉 MCP Server POC is ready!
The server can now be used with MCP clients to scrape websites with JS disabled.

🔌 Connecting to MCP Clients

Claude Desktop Integration

1. Find your Claude Desktop config file:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
   • Linux: ~/.config/Claude/claude_desktop_config.json

2. Add the server configuration:

{
  "mcpServers": {
    "camoufox-scraper": {
      "command": "poetry",
      "args": ["run", "python", "/full/path/to/your/mcp-camoufox-scraper/run_server.py"],
      "cwd": "/full/path/to/your/mcp-camoufox-scraper"
    }
  }
}

3. Replace the path with your actual project directory:

# Get your full path
pwd
# Copy the output and use it in the config above

4. Restart Claude Desktop completely (quit and reopen)

5. Verify connection by asking Claude:

   "What MCP servers do you have access to?"

You should see "camoufox-scraper" listed with the available tools.

Other MCP Clients

For other MCP clients, use this server configuration:

• Command: poetry
• Arguments: ["run", "python", "/path/to/run_server.py"]
• Working Directory: /path/to/mcp-camoufox-scraper
• Communication: stdio

🛠️ Features

• Dual JavaScript mode: JavaScript enabled for network monitoring, disabled for clean HTML extraction
• Network request monitoring: Capture all XHR/API calls and HTTP requests made during page load
• Clean HTML extraction: Get HTML content with JavaScript disabled to avoid dynamic modifications
• MCP protocol integration: Works with any MCP client (Claude Desktop, etc.)

📋 Available Tools

1. navigate_to_url

Navigate to a URL with JavaScript enabled to capture network requests and dynamic content.

Parameters:

• url (required): The URL to navigate to
• wait_time (optional): Time to wait after page load in seconds (default: 3)

Claude Example:

"Please navigate to https://example.com and wait 5 seconds"

2. get_page_html

Extract clean HTML content by re-loading the page with JavaScript disabled.

Parameters: None

Claude Example:

"Get the HTML content from the current page"

3. get_network_requests

Get all captured network requests from the last page navigation.

Parameters:

• filter_type (optional): Filter by request type ("xhr", "fetch", "all") - default: "all"

Claude Example:

"Show me all the network requests that were captured"

4. close_browser

Close the browser and cleanup resources.

Parameters: None

Claude Example:

"Close the browser to free up resources"

💡 Usage Examples

Basic Web Scraping

You: Navigate to https://news.ycombinator.com
Claude: [Uses navigate_to_url tool]
You: Get the HTML content
Claude: [Uses get_page_html tool and analyzes the content]
You: What network requests were made?
Claude: [Uses get_network_requests tool and shows API calls]

API Discovery

You: Go to https://httpbin.org/headers and show me what requests it makes
Claude: [Navigates and shows network monitoring results]

Content Analysis

You: Navigate to https://example.com and get both the network requests and clean HTML
Claude: [Navigates with JS enabled to capture requests, then extracts HTML with JS disabled]
You: What's the difference between the two modes?
Claude: [Explains that navigation captures dynamic requests while HTML extraction gives clean content]

🧪 Testing & Verification

Run Full Test Suite

poetry run python test_mcp_server.py

Expected output:

=== Testing MCP Server Tools ===
✓ Navigation successful: Example Domain
✓ HTML extraction successful
✓ Network requests retrieval successful
✓ Complex site navigation successful
✓ Browser closed successfully
🎉 MCP Server POC is ready!

Test Individual Components

# Test just the Camoufox API
poetry run python test_camoufox_api.py

# Start server manually (for debugging)
poetry run python run_server.py

🔧 Troubleshooting

Quick Diagnosis

# Run the setup verification script first
poetry run python setup_verify.py

This will check your Python version, dependencies, project structure, and generate the correct Claude Desktop configuration.

Server Won't Start

# Check if dependencies are installed
poetry show mcp camoufox

# Run the test to identify issues
poetry run python test_mcp_server.py

Claude Desktop Connection Issues

1. Check config file location - Make sure you're editing the right file
2. Use absolute paths - Relative paths won't work
3. Restart Claude Desktop completely after config changes
4. Check Claude's developer tools for error messages
5. Verify Python path - Make sure python command works in terminal

Browser Issues

• Camoufox download: First run may take time downloading browser binaries
• Permission errors: Make sure the script has execute permissions (chmod +x run_server.py)
• Port conflicts: Close other browser automation tools if running

Common Error Messages

• "No such file or directory" → Check the path in your config
• "Permission denied" → Run chmod +x run_server.py
• "Module not found" → Run poetry install

📁 Project Structure

mcp-camoufox-scraper/
├── mcp_camoufox_scraper/
│   ├── __init__.py
│   └── server.py              # Main MCP server implementation
├── run_server.py              # CLI runner script  
├── setup_verify.py            # Setup verification & config generator
├── test_mcp_server.py         # Comprehensive test suite
├── test_camoufox_api.py       # Camoufox API validation
├── pyproject.toml            # Project configuration & dependencies
├── poetry.lock               # Locked dependency versions
└── README.md                 # This file

⚠️ Limitations

• Dual-context approach: HTML extraction requires re-navigation with JS disabled
• Single page: Only one page can be active at a time
• No authentication: Currently no support for login/auth workflows
• Limited to HTTP/HTTPS: No support for other protocols

🔒 Security & Privacy

• No data collection: Server runs locally, no data sent to external services
• Controlled JavaScript execution: JS enabled only when needed for network monitoring
• Privacy-focused browser: Camoufox is designed for privacy
• Local execution: All scraping happens on your machine

🛠️ Development & Extension

Modifying the Server

1. Edit mcp_camoufox_scraper/server.py
2. Test changes: poetry run python test_mcp_server.py
3. Verify MCP compliance with your client

Adding New Tools

# In server.py, add to _register_tools():
@self.server.call_tool()
async def handle_call_tool(name: str, arguments: Dict[str, Any]):
    if name == "your_new_tool":
        return await self._your_new_tool(arguments)

Technical Details

• Browser: Camoufox (privacy-focused Firefox-based)
• Protocol: MCP (Model Context Protocol)
• Language: Python 3.10+ with asyncio
• Dependencies: MCP SDK, Camoufox browser automation

📝 License

MIT License - Feel free to use, modify, and distribute.

🤝 Contributing

This is a proof-of-concept project. Feel free to:

• Fork and extend for your use cases
• Submit issues and improvements
• Share your modifications with the community

---

Happy scraping! 🕷️