Web Crawler MCP

A powerful web crawling tool that integrates with AI assistants via the MCP (Machine Conversation Protocol). This project allows you to crawl websites and save their content [...]

📋 Features

• Website crawling with configurable depth
• Support for internal and external links
• Generation of structured Markdown files
• Native integration with AI assistants via MCP
• Detailed crawl result statistics
• Error and not found page handling

🚀 Installation

Prerequisites

• Python 3.9 or higher

Installation Steps

1. Clone this repository:

git clone laurentvv/crawl4ai-mcp
cd crawl4ai-mcp

2. Create and activate a virtual environment:

# Windows
python -m venv .venv
.venv\Scripts\activate

# Linux/MacOS
python -m venv .venv
source .venv/bin/activate

3. Install the required dependencies:

pip install -r requirements.txt

🔧 Configuration

MCP Configuration for AI Assistants

To use this crawler with AI assistants like VScode Cline, configure your cline_mcp_settings.json file:

{
  "mcpServers": {
    "crawl": {
      "command": "PATH\\TO\\YOUR\\ENVIRONMENT\\.venv\\Scripts\\python.exe",
      "args": [
        "PATH\\TO\\YOUR\\PROJECT\\crawl_mcp.py"
      ],
      "disabled": false,
      "autoApprove": [],
      "timeout": 600
    }
  }
}

Replace PATH\\TO\\YOUR\\ENVIRONMENT and PATH\\TO\\YOUR\\PROJECT with the appropriate paths on your system.

Concrete Example (Windows)

{
  "mcpServers": {
    "crawl": {
      "command": "C:\\Python\\crawl4ai-mcp\\.venv\\Scripts\\python.exe",
      "args": [
        "D:\\Python\\crawl4ai-mcp\\crawl_mcp.py"
      ],
      "disabled": false,
      "autoApprove": [],
      "timeout": 600
    }
  }
}

🖥️ Usage

Usage with an AI Assistant (via MCP)

Once configured in your AI assistant, you can use the crawler by asking the assistant to perform a crawl using the following syntax:

Can you crawl the website https://example.com with a depth of 2?

The assistant will use the MCP protocol to run the crawling tool with the specified parameters.

Usage Examples with Claude

Here are examples of requests you can make to Claude after configuring the MCP tool:

• Simple Crawl: "Can you crawl the site example.com and give me a summary?"
• Crawl with Options: "Can you crawl https://example.com with a depth of 3 and include external links?"
• Crawl with Custom Output: "Can you crawl the blog example.com and save the results in a file named 'blog_analysis.md'?"

📁 Result Structure

Crawl results are saved in the crawl_results folder at the root of the project. Each result file is in Markdown format with the following structure:

# https://example.com/page

## Metadata
- Depth: 1
- Timestamp: 2023-07-01T12:34:56

## Content
Extracted content from the page...

---

🛠️ Available Parameters

The crawl tool accepts the following parameters:

Parameter	Type	Description	Default Value
url	string	URL to crawl (required)	-
max_depth	integer	Maximum crawling depth	2
include_external	boolean	Include external links	false
verbose	boolean	Enable detailed output	true
output_file	string	Output file path	automatically generated

📊 Result Format

The tool returns a summary with:

• URL crawled
• Path to the generated file
• Duration of the crawl
• Statistics about processed pages (successful, failed, not found, access forbidden)

Results are saved in the crawl_results directory of your project.

🤝 Contribution

Contributions are welcome! Feel free to open an issue or submit a pull request.

📄 License

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