Skip to main content
Web fetch reads a specific URL you name into the model’s context while it answers. Where web search starts from a query and goes looking for sources, web fetch starts from a URL that is already in your prompt and pulls that one page in. The everyday case: you paste a link and ask “summarize this,” “what changed in this release,” or “is this relevant to what I am building.” Without fetch, the model can only guess from the URL text. With fetch, it reads the actual page and answers from its real contents.

​
Turning On Web Fetch

Web fetch is controlled by the --web-fetch flag on your ask command. It takes on or off, and the default is on. Fetch runs when your prompt contains an explicit URL. With no URL in the prompt, there is nothing to fetch.
You can also run this from your terminal with nito ask --web-fetch on "Fetch https://example.com and tell me the main heading text.". See Ask and Fusion from the terminal.

​
Worked Example

Run the command above and Nito retrieves the page, then reports its main heading text: Example Domain. The answer comes from the page’s real contents, not from the model guessing based on the URL. For everyday use, paste any link and ask a question about it:
They solve different problems, and you pick based on what you start with. If your prompt names a page, reach for web fetch. If it asks a question that needs Nito to go discover sources, reach for web search.

​
Notes

  • Fetch runs only on explicit URLs. If there is no link in your message, there is nothing to fetch; use web search for query-driven retrieval instead.
  • Fetch reads public web content. Do not rely on it for authenticated or internal pages.
  • Web fetch works with ask and inside a session, and in a Fusion all participants share the same web setting.

Web Search

Query-driven retrieval when you do not have a specific link.

Server-Side Tools

How Nito’s built-in tools work in general.