Search & Web

Fetch

Retrieves a URL's content and converts it to clean markdown for an agent to read, with chunked pagination for long pages.

Publisher

Model Context Protocol project

View repo ↗

Install

uvx mcp-server-fetch

The Fetch server retrieves a given URL and converts its HTML into clean markdown sized for an LLM context window, with a start-index parameter so an agent can page through long documents in chunks. It’s one of the official reference servers in the modelcontextprotocol/servers repo.

Why it matters

It gives an agent a simple, dependable way to read a specific webpage on demand — a targeted complement to a general search server like Brave Search.

Good for

Pulling in a specific doc page, changelog, or article an agent has already been pointed to, rather than open-ended web search.