Web Fetch (web_fetch)
Fetches content from a URL, converts HTML to Markdown, and processes it with a focused AI call.
Parameters
| Parameter | Required | Description |
|---|---|---|
url | Yes | Fully-formed URL (http:// or https://) |
prompt | Yes | Describe what to extract from the page |
Behaviour
- HTML is converted to readable text before processing.
- GitHub blob URLs are automatically converted to raw URLs.
- HTTP URLs are upgraded to HTTPS.
- Processes one URL per call.
Confirmation
Requires confirmation before fetching.
Examples
web_fetch(url="https://example.com/blog/post", prompt="Summarise the main points")
web_fetch(url="https://github.com/org/repo/blob/main/README.md", prompt="What are the installation steps?")MCP alternative
If an MCP-provided web fetch tool is available (any tool starting with mcp__), prefer it — it may have fewer restrictions than the built-in tool.
Last updated on