platform/seed/tools/web-fetch/TOOL.md

---

name: web_fetch label: Web Fetch description: Fetch and return the text content of a URL. Tries to get clean readable content using a cascade of strategies in order: appending .md to the URL, fetching a /llms.txt discovery file, requesting plain text via Accept header, then falling back to stripping HTML. Use when the user provides a URL and wants to read, summarize, or extract information from web content. language: typescript inputs: url: type: string description: The URL to fetch content from

Web Fetch

Fetches web content with cascading fallback strategies to get the cleanest possible text.

Strategies (in order)

1. Markdown version — Appends .md to the URL (works on GitHub, many docs sites)
2. llms.txt discovery — Checks /llms.txt at the root (sites that publish LLM-friendly content)
3. Plain text request — Sends Accept: text/plain header
4. HTML strip — Fetches HTML and strips tags, scripts, and styles

Output

Returns the text content with a note indicating which strategy succeeded.