Fetch
Retrieves a URL and converts the page to markdown so an agent can read it without burning tokens on raw HTML.
What it does
Fetches a web page and converts it to markdown before handing it to the model. Raw HTML is mostly tags and script, so converting first can cut the token cost of reading a page substantially.
An official reference server, written in Python.
Setup
claude mcp add fetch -- uvx mcp-server-fetch
Requires uv. There is a pip route too:
pip install mcp-server-fetch python -m mcp_server_fetch
What to watch out for
It fetches whatever URL the model asks for, which is a prompt injection surface. A page can contain text designed to instruct your agent. Treat fetched content as untrusted input, not as instructions.
It does not run JavaScript, so single page apps that render client side will come back mostly empty. Use a browser automation server such as Playwright for those.
Respect robots.txt and rate limits on sites you do not own.
