Engineering deep-dives, product updates, and practical guides for building agents on web data.
How the Read engine strips boilerplate and preserves structure — deterministically, with no LLM in the loop.
Map a site, read every page as markdown, chunk, embed — always-fresh retrieval in a handful of calls.
Two endpoints, two jobs. Map discovers URLs for 1 credit; Crawl reads them. Here's how to combine them.
One config and Claude, Cursor, or ChatGPT can read, crawl, and extract — as native tools.
Resolve a domain into name, logo, colors, fonts, socials, and address — read live from the site.
A static read and an AI extraction don't cost the same to run — so they shouldn't cost the same.