Fetch MCP Server

Official

Fetch any web page or API endpoint and pass the content to your AI

fetchhttpwebofficialmarkdownapi

The official Fetch MCP server lets your AI assistant retrieve content from any URL. It fetches web pages, converts HTML to clean markdown, and handles JSON APIs — all through a simple interface. Supports custom headers for authenticated requests. Lighter and faster than the Puppeteer server for pages that don't require JavaScript rendering. Perfect for reading documentation, fetching API responses, and giving the AI access to live web content.

Install Config

Config file: ~/Library/Application Support/Claude/claude_desktop_config.json· Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
      "args": [
        "mcp-server-fetch"
      ]
    }
  }
}

Available Tools (1)

Tool name	Description
fetch	Fetch a URL and return the content as text or markdown

Example Prompts

“Fetch the React documentation for useEffect and summarize it”
“Get the current weather JSON from this API endpoint”
“Read the README from this GitHub repository URL”
“Fetch this product page and extract the price”
“What does the Hacker News front page say today?”

Pros & Cons

Pros

Simple — one tool, fetch any URL
Converts HTML to clean markdown automatically
No browser required — fast and lightweight
Official Anthropic server

Cons

Cannot execute JavaScript — static HTML only
Cannot interact with pages (click, fill forms)
Blocked by sites that require authentication cookies

Who Is This For?

·Developers who want the AI to read documentation or API responses
·Researchers who want to quickly summarize web pages

When to use: Use Fetch for any URL that returns readable content — documentation, blog posts, APIs, and static pages. Use Puppeteer instead when the page requires JavaScript to render content.