Fetch
Fetch a URL and return clean Markdown — the canonical web-reading tool.
Install
Add this command to your MCP client config to connect Fetch.
uvx mcp-server-fetchAbout Fetch
The fetch server takes a URL, downloads it, strips the boilerplate, and returns Markdown. It handles HTML→Markdown conversion (so the model gets readable content rather than nav menus and cookie banners) and respects robots.txt by default.
It is the simplest and most-used MCP server in practice — any time you say "go read this page and summarize it," fetch is the thing actually doing the reading. Pair it with brave-search or tavily for "search the web, then read the top 3 results" workflows.
Tags
Related servers
Other MCP servers in Development and beyond.
Filesystem
Anthropic
Read, write, and search local files with sandboxed directory access.
Git
Anthropic
Read, search, and manipulate local Git repositories.
GitLab
Anthropic
Issues, merge requests, and project management for GitLab.com and self-hosted.
Puppeteer
Anthropic
Drive a real Chromium browser — navigate, click, fill forms, screenshot.
GitHub
GitHub
Issues, PRs, code search, and repository operations across GitHub.
Playwright
Microsoft
Cross-browser automation — the same Playwright you write tests with.