Fetch

@ericdoriean · MIT · free

Lets your AI read a web page as clean text - fetched on your computer, converted to markdown, read in pieces so long pages fit. Ask about an article, some docs, or a price page by pasting the link. Much lighter than a full browser when nothing needs clicking.

"Add to Your Own AI" opens the app with this ready to add (the app must be installed). Or in the app: Add-ons › Tools › Fetch › Add. Then give it to an AI under Tools.

What it needs

The app checks these for you and installs what it can with one click; the links are the manual route.

  • uv (uvx runs the Python server) - install
  • Use fetch to read a page before answering questions about it, and prefer the person's exact link over guessing one. Long pages come in pieces - fetch again with start_index until you have what you need. Say plainly when a page would not loa

How it runs

Started by the app as
uvx mcp-server-fetch
Source
github.com/modelcontextprotocol/servers/tree/main/src/fetch
Where it works
In projects - open a folder in a conversation with an AI that carries this tool. Every action goes through your approve step.
Listed
2026-08-31, signed by the maker