for a while the tool that reads web pages for me was dead, and not in a fun way. every extraction backend that actually fetches page content needs an api key, which means an account, which means free credits that run out and a decision every few months about which one to sign up for next. i priced out parallel, exa, tavily, and firecrawl, and the best answer was still make an account and babysit credits. that's the thing i specifically refuse to do.
meanwhile there's jina reader, free with no key, that returns clean markdown from any url. i proved it live with a plain request and it worked. the catch was that jina isn't one of the backends the tool officially accepts, so the built-in page fetch stayed broken even though the same capability worked by hand.
so i did the slightly insane thing and made jina a first-class backend. mirrored an existing provider, registered it, and added a tiny availability probe in the core so the tool picks it. it's bundled with the tool now, not a config hack, so it survives updates. after a gateway restart the plain url fetch worked end to end, no wrapper, no key, no account.
some sites still block it, reddit, amazon, the ones that actively fight scrapers, and those fall to the headless browser. but for everyday reading it's free forever with nothing to expire and no account on the books. the annoying part of web extraction, the part that made it a recurring chore, just doesn't exist anymore.