@upcrawl/mcp

MCP (Model Context Protocol) server for Upcrawl - Web scraping and search for AI agents.

Give Claude, Cursor, Windsurf, and other MCP-compatible AI assistants the ability to scrape web pages and search the internet.

Quick Start

npx -y @upcrawl/mcp

That's it. On first run it will:

1. Open your browser to grab your API key ($10 free credits on signup)
2. Ask you to paste the key
3. Ask which client to configure (Claude Code, Claude Desktop, Cursor, VS Code, Windsurf)
4. Set everything up automatically

Manual Setup

If you prefer to set things up manually:

Claude Code

claude mcp add upcrawl -- npx -y @upcrawl/mcp

Claude Desktop

Add to ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "upcrawl": {
      "command": "npx",
      "args": ["-y", "@upcrawl/mcp"]
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "upcrawl": {
      "command": "npx",
      "args": ["-y", "@upcrawl/mcp"]
    }
  }
}

VS Code (Copilot)

Add to .vscode/mcp.json in your project root:

{
  "servers": {
    "upcrawl": {
      "command": "npx",
      "args": ["-y", "@upcrawl/mcp"]
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "upcrawl": {
      "command": "npx",
      "args": ["-y", "@upcrawl/mcp"]
    }
  }
}

For manual setup, you'll need an API key. Pass it via --api-key=uc-xxx arg, UPCRAWL_API_KEY env var, or run npx -y @upcrawl/mcp once to save it.

Key priority: --api-key flag > UPCRAWL_API_KEY env var > ~/.upcrawl/config.json

Available Tools

scrape

Scrape content from one or more URLs with LLM-powered summarization.

Parameters:

• urls (required): Array of URLs to scrape (one or more)
• query (required): Instruction for how to process the scraped content
• type: Output format - markdown (default) or html
• onlyMainContent: Extract only main content (default: true)

Example usage:

"Scrape https://example.com and extract the main points" "Scrape these three docs pages and compare their APIs: [url1, url2, url3]"

search

Search the web and get results with URLs, titles, and descriptions.

Parameters:

• query (required): The search query
• limit: Number of results (1-100, default: 10)
• location: Location for results (e.g., "US", "IN")

Example usage:

"Search for the latest AI agent frameworks"

Development

# Clone the repo
git clone https://github.com/upcrawl/upcrawl-mcp
cd upcrawl-mcp

# Install dependencies
npm install

# Build
npm run build

# Run locally
UPCRAWL_API_KEY=your-key node dist/index.js

License

MIT