modelcontextprotocol/server-fetch

Efficient web content fetching and processing MCP server, designed for AI consumption and integration with the MCP framework.

Source on GitHub

Features

• Web Content Fetching: Fetches content from any provided URL.
• HTML to Markdown Conversion: Converts fetched HTML content to markdown for easier processing by LLMs and AI systems.
• Response Truncation and Pagination: Supports truncating responses and specifying a start_index to read webpages in chunks, allowing for incremental retrieval until all desired information is obtained.
• Flexible Output: Option to return raw HTML content instead of markdown.
• Configurable via Arguments:
  • url (required): The URL to fetch.
  • max_length (optional): Maximum number of characters to return (default: 5000).
  • start_index (optional): Character index to start content extraction (default: 0).
  • raw (optional): Whether to return raw HTML (default: false).
• Multiple Installation Options:
  • Via uvx (no installation needed, direct run).
  • Via pip (pip install mcp-server-fetch).
  • Via Docker.
• Integration Ready: Easily configurable for use with Claude.app, VS Code, and other MCP-compatible applications.
• robots.txt Compliance: Obeys robots.txt by default for model tool requests; can be disabled via configuration.
• User-Agent Customization: Allows specifying a custom user-agent string.
• Proxy Support: Can be configured to use a proxy with the --proxy-url argument.
• Debugging Tools: Compatible with the MCP inspector for debugging and development.
• Open Source and Extensible: Licensed under MIT, open for contributions, and extensible with new tools and enhancements.

Pricing

No pricing information provided. The server is open source and available under the MIT License.

Tags

mcp, web, fetch, ai-integration