Fetch is a Model Context Protocol (MCP) server. It retrieves web content and converts HTML to markdown, making it easier for Large Language Models to process and use information from web pages.
To start using Fetch, install it via node.js or pip. Run the server with the command python -m mcp_server_fetch or via uvx as shown in the docs. To fetch content, simply call the fetch tool with a URL.
Its key features include fetching URLs and extracting content as markdown. It supports configuration like maximum content length and extraction start index. It also allows custom user-agent strings and respects robots.txt rules.
Use cases include enabling LLMs to access web data for various apps, converting online articles into a simplified format for analysis, and assisting in data retrieval for research or aggregation workflows.
For common questions: Fetch can extract content from most web pages, but results depend on a site's structure and restrictions. It is designed for easy integration with LLMs and can be customized. While Fetch itself has no strict usage limits, your implementation must follow the guidelines of the websites you access.
{
"mcpServers": {
"fetch": {
"args": [
"mcp-server-fetch"
],
"command": "uvx"
}
}
}