Reader

Reader is a web-context API aimed at agents that need a page, not a browser farm. POST https://api.reader.dev/v1/read with a URL returns cleaned Markdown. The same product also crawls from a start URL with depth and pattern limits, extracts JSON from a schema or instruction, and can open a cloud browser session for Playwright or Puppeteer when the page needs clicks.

Nihal Kaul's own write-up (reader.dev/blog/why-i-built-reader) is the build note: every agent-plus-web project rebuilt the same stack — JS rendering, main-content extract, proxy rotation, anti-bot — so the open-source engine and the hosted API are that stack, once. The README cites Ulixee Hero under the hood for stealth browsing and a recycled browser pool. The cloud plan advertises 1,000 free credits a month with no card.

Useful if you are stuffing docs or product pages into a RAG pipeline and do not want to own Puppeteer. It is infrastructure, not a consumer app. The live site, console, and GitHub repo all resolved on this check.