Crawl4AI MCP server: web access for Claude and Cursor
The official Crawl4AI Docker image includes an MCP server. Start the container, add http://localhost:11235/mcp/sse to your client, and your AI assistant can fetch pages as Markdown, take screenshots and run crawls.
Updated 6 Nov 2026
1. Start the Crawl4AI Docker server
docker run -d -p 11235:11235 --name crawl4ai --shm-size=1g unclecode/crawl4ai:latestThe server then listens on http://localhost:11235. Two MCP transports are available:
- SSE:
http://localhost:11235/mcp/sse - WebSocket:
ws://localhost:11235/mcp/ws
2. Connect your client
Claude Code
claude mcp add --transport sse c4ai-sse http://localhost:11235/mcp/sse
claude mcp listCursor (and other clients with URL support)
{
"mcpServers": {
"crawl4ai": { "url": "http://localhost:11235/mcp/sse" }
}
}Claude Desktop
Claude Desktop starts MCP servers as local commands. A common way to reach an SSE server is the mcp-remote bridge:
{
"mcpServers": {
"crawl4ai": { "command": "npx", "args": ["mcp-remote", "http://localhost:11235/mcp/sse"] }
}
}3. Tools your assistant gets
| Tool | What it does |
|---|---|
md | Fetch a URL as Markdown (raw, fit or filtered by a query) |
crawl | Run a full crawl with browser and crawler configuration |
html | Return cleaned HTML |
screenshot / pdf | Capture the page as image or PDF |
execute_js | Run JavaScript on the page and return the result |
ask | Query the Crawl4AI library context (documentation help) |
The exact schemas are listed at http://localhost:11235/mcp/schema.
Crawl4AI RAG MCP server
“Crawl4AI RAG MCP” usually refers to a separate community project (mcp-crawl4ai-rag by coleam00) that crawls pages with Crawl4AI and stores them in a vector database, so the assistant can search them later. Use the official Docker MCP for live fetching, a RAG MCP when you want a searchable knowledge base.
config.yml before exposing it.No Docker at hand? Test a page in the online tool first.
Crawl4AI MCP FAQ
Does Crawl4AI have an MCP server?
How do I add Crawl4AI to Claude Code?
claude mcp add --transport sse c4ai-sse http://localhost:11235/mcp/sse.