furl
You want an AI agent to read a documentation page, so you reach for curl. These are the first two lines it gives you back:
curl -s https://bun.sh/docs/installation | head -2<!DOCTYPE html><html lang="en" class="inter_1d81deff-module__CYM0aG__variable papermono…
function a(p){if(p==null)return"/";p=""+p;if(""===p)return"/";return"/"===p.charAt(0)?p…Minified JavaScript. The prose you actually wanted is somewhere in the other half-megabyte, wrapped in nav chrome, script tags, and inline styles. Whatever reads this next — an agent, or you — pays for all of it.
furl fetches the same URL and prints the page as markdown:
furl https://bun.sh/docs/installation | head -6> ## Documentation Index
> Fetch the complete documentation index at: https://bun.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# InstallationSame page, 522 KB of HTML down to 13 KB of markdown, and every byte of it is content.
No key, no signup
Most fetches need neither. furl asks the origin for its own markdown before it considers converting anything, and a surprising number of sites — documentation sites especially — already publish one. Only when that fails does furl fall back to a conversion provider, and the default one is keyless.
What it isn't
furl fetches one URL and prints markdown. It isn't a crawler, a scraper framework, or a headless browser. There's no caching, so the same URL hits the network every time, and no timeout, so an unresponsive origin will hang until you interrupt it.
Where to go next
- Getting started — install furl and run your first fetch
- Fetching pages — flags, output streams, and exit codes
- How furl finds markdown — the four steps behind a fetch, and one trap worth knowing
- Providers — jina, exa, firecrawl, and where API keys live