Skip to main content

​
What it does

Fetches a URL via HTTP and converts the response to clean text:
  • HTML — converted to readable text (headings, lists, links preserved)
  • JSON — pretty-printed
  • Plain text — returned as-is
Much faster and cheaper than launching a browser. The assistant prefers this for server-rendered pages and direct URLs.

​
Requirements

None. Web fetch is always available — no configuration needed.

​
How the assistant uses it

When the assistant needs to read a web page and doesn’t need to interact with it (no clicking, no JavaScript rendering), it calls web_fetch with a URL. Common scenarios:
  • Reading documentation pages
  • Fetching API responses
  • Reading articles or blog posts
  • Checking content at a URL you shared in chat

​
Configuration

​
Limitations

  • No JavaScript execution — won’t work for SPAs or client-rendered pages (use the browser tool instead)
  • No authentication — can’t access pages behind login
  • Follows redirects automatically
  • Truncates output at maxChars to stay within LLM context limits
If the assistant says it can’t read a page or gets empty content, the page likely requires JavaScript. Ask it to use the browser tool instead.