概览
What is Crawl4AI MCP Server?
Crawl4AI MCP Server is an MCP (Model Context Protocol) server that equips AI assistant systems with multi‑engine search and LLM‑optimized web page content understanding. It supports DuckDuckGo (default) and Google search, and provides intelligent content extraction with citation‑enabled output formats.
How to use Crawl4AI MCP Server?
Install via cloning the repository, creating a Python 3.9+ virtual environment, installing dependencies (pip install -r requirements.txt) and the Playwright browser (playwright install), or via Smithery for Claude Desktop. The server exposes two tools: search (with parameters query, num_results, engine) and read_url (with parameters url and format). To use Google search, configure config.json with an API key and CSE ID.
Key features of Crawl4AI MCP Server
- Multi‑engine search (DuckDuckGo and Google)
- LLM‑optimized web content extraction with noise filtering
- Automatic identification and retention of key content
- Multiple output formats including markdown with citations
- High‑performance asynchronous design based on FastMCP
Use cases of Crawl4AI MCP Server
- AI assistants performing web research and fact‑checking
- Summarizing and extracting core information from articles
- Gathering search results from multiple engines simultaneously
- Providing LLMs with clean, citation‑tracked web content
FAQ from Crawl4AI MCP Server
Which search engines are supported?
DuckDuckGo (default, no API key required) and Google (requires API key and CSE ID configured in config.json). The "engine" parameter can be set to "duckduckgo", "google", or "all".
What Python version is required?
Python 3.9 or higher is required.
What output formats does the read_url tool support?
It supports markdown_with_citations (default), fit_markdown, raw_markdown, references_markdown, fit_html, and markdown.
Is an API key needed to use DuckDuckGo search?
No, DuckDuckGo search works without any API key. Google search requires a Google API key and a Custom Search Engine ID.
What dependencies are needed beyond Python?
The Playwright browser must be installed (playwright install). All other dependencies are listed in requirements.txt.
