Fetch MCP

What it does

Fetch MCP is a lightweight HTTP client for Claude. It makes standard GET (and optionally POST) requests to any URL and returns the response content in multiple formats: raw HTML, cleaned Markdown, or parsed JSON. No browser required — no JavaScript execution.

This makes it ideal for fetching documentation pages, API responses, RSS feeds, configuration files hosted on the web, or any static content. It is significantly faster than browser-based alternatives like Playwright because it skips the rendering engine entirely.

• Fetch any public URL via HTTP/HTTPS
• Return page content as clean Markdown (stripped of boilerplate)
• Return raw HTML for parsing or analysis
• Fetch and parse JSON APIs directly
• Respect robots.txt and rate limiting

How to connect

claude mcp add fetch -- npx -y @modelcontextprotocol/server-fetch

Available tools

• fetch — retrieve a URL, returning raw content
• fetch_markdown — fetch a URL and convert HTML to Markdown
• fetch_html — fetch a URL and return raw HTML
• fetch_json — fetch a URL, parse and return JSON

Example usage

Ask Claude to summarize a documentation page, pull data from a public API, or read an RSS feed:

# User prompt
Fetch the Stripe API changelog at
stripe.com/docs/changelog and summarize
the last 5 changes.

fetch_markdown("https://stripe.com/docs/changelog")
→ returns ~8,000 tokens of markdown
→ Claude extracts 5 most recent entries
→ formats as bullet summary