Purpose
Crawleo-MCP is Crawleo's Model Context Protocol (MCP) server, providing AI assistants with real-time web search and crawling capabilities. Through native tool integration, AI assistants such as Claude can directly access real-time web data.
Main Capabilities
- Real-time web search: Supports any country/language.
- Multiple output formats: Enhanced HTML, raw HTML, Markdown, plain text.
- Device-specific results: Desktop, mobile, or tablet views.
- Deep content extraction: Supports JavaScript rendering.
- Zero data retention: Complete privacy protection.
- Automatic crawling: Optionally automatically crawl search results.
Usage
Installation Options
- NPM (recommended for local use): Install globally or use npx, no installation required.
- Clone the repository.
- Docker: Build and run with Docker, and configure the MCP client.
- Remote server (no installation required): Use the hosted version https://api.crawleo.dev/mcp。
Get an API Key
Visit crawleo.dev, register for a free account, and copy the API key starting with sk_ from the dashboard.
Configuration
- Claude Desktop / Cursor / Windsurf (local): Add the Crawleo MCP configuration to the corresponding configuration file, replacing YOUR_API_KEY_HERE with the actual key.
- macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows: %APPDATA%\Claude\claude_desktop_config.json
- Linux: ~/.config/Claude/claude_desktop_config.json
- Remote server (hosted): Also configure it in the configuration file.
- GitHub Copilot: Configure the MCP server in VS Code or a compatible editor.
- OpenAI Platform (direct integration): Use parameters such as server_url, authorization, allowed_tools, require_approval.
Available Tools
- web.search: Search the web in real time, with customizable parameters.
- query (required)
- max_pages (number of result pages, default 1)
- setLang (language code)
- cc (country code)
- device (device type: desktop, mobile, tablet, default desktop)
- enhanced_html (default true)
- raw_html (default false)
- markdown (default true)
- page_text (default false)
- auto_crawling (default false)
- web.crawl: Extract content from a specified URL.
- urls (required)
- rawHtml (default false)
- markdown (default false)
- screenshot (optional)
- country (geolocation)
Use Cases
- Research assistant
- Content analysis
- Code documentation
- News monitoring
- Market research
Limitations and Notes
- API keys start with sk_ and must be placed in quotes; Claude/Cursor/Windsurf must include the "Bearer " prefix in the Authorization header, while the OpenAI Platform uses the key directly.
- Tool names must be correct: use web.search (not search_web), web.crawl (not crawl_web).
- If you encounter issues, check the configuration file location, JSON syntax, restart the application, and verify the API key and account balance.
- Pricing is the same as the API: 10,000 searches → $20; 100,000 searches → $100; 250,000 searches → $200.
Privacy and Security
- Zero data retention: Search queries or results are not stored.
- Secure authentication: API keys are transmitted via HTTPS.
- No tracking: Usage patterns remain private.
