what this is

deeprat is a crawler that lives in your browser tab. when you ask something, eight rats fan out and fetch real pages over the network — you can watch every request land in the log, with its size and timing. nothing here is a replay.

where the rats can go

• en.wikipedia.org — search + full text extracts + link maps, via the public API
• hn.algolia.com + hacker-news.firebaseio.com — hacker news search and items
• export.arxiv.org — the rats try; arxiv often refuses browser tabs, and when it does the log shows an honest ERR row instead of invented data

these hosts allow browser tabs in (CORS). most of the web doesn't — that's why rats, not a proxy. nothing routes through any server of ours; there isn't one.

the answer rule

rats don't summarize and don't make anything up. the answer pane shows the exact sentences they bit off the fetched pages, ranked by how many of your words each sentence contains, each with a [n] pointing at its source. if the rats come back with nothing readable, the answer says so.

reading the panes

• hop pane — the page a rat is holding. space cycles through fetched pages.
• tail -f — the live request feed: rat, GET, url, kB, ms. red = error.
• bottom table — the eight processes, htop style: state, pages, time, current url.

$DEEPRAT

the token behind the rats, on Solana. one official CA — it lives in the box on the live tab and in the top bar, nowhere else. the mcap chip and chart link go live from Dexscreener once the pair exists; until then they read "—" instead of pretending.

privacy

your questions leave this tab only as search queries to the hosts above. no account, no cookies, no backend. close the tab and the session is gone.