pplx — What the web says, ranked and quoted — openbash

Perplexity's index from the shell: ranked results for a question, and the passages of a page that bear on it.

pplx

pplx is a command on the openbash shell, reachable over HTTP, SSH and MCP with one key. It composes with the other programs through pipes; there is no per-program endpoint and no schema for the join.

pplx search web "who acquired Wiz" -n 5 --recency-filter month

help

pplx: the live web, searched and quoted, as JSON.

USAGE
  pplx search web       <QUERY> [MORE PHRASINGS…] [-n N] [--domains …] [--recency-filter W]
  pplx content snippets <QUERY> <URL> [URL…]      [--max-tokens N] [--max-tokens-per-page N]

  Several phrasings after `search web` are ONE search, ranked together — the
  way to ask about a thing whose name you are not sure of.

NOTHING IS QUOTED OUT OF A PAGE THE ACCOUNT DOES NOT HOLD
  `search web` returns addresses and dates and withholds the engine's snippet,
  as `browser search` does: a snippet is enough to lift a figure out of and not
  enough to have read the page it came from.

  `content snippets` quotes, and earns it — every url is TAKEN first, through
  `browser asset`, so each page is in the account's store under the address a
  citation carries before a word of it is printed. `cat`, `pdftext` and a
  citation all reach it by that same url afterwards.

  What is held is the SOURCE, not the sentence. Perplexity quotes its own crawl
  and this takes its own copy moments later; a paywall or a re-render can make
  the two differ. The guarantee is that the page is in the store and checkable.

RETURNS
  One JSON object on stdout. `search web` answers with `results`, each carrying
  a title, a url and its dates. `content snippets` answers with one entry per
  url — the passages of that page bearing on the query, with `…` where it
  elided — plus `not_taken`, the addresses that could not be brought into the
  store and were therefore not quoted. A page taken but unreadable by the index
  carries an `error` of its own without failing the others.

COSTS A FETCH PER PAGE
  `content snippets` is one `browser asset` per url on top of the index call,
  billed as `web_scrape` — free for a url the account already holds, and a
  provisioned browser where an origin blocks the direct fetch.

DOES NOT COVER
  Anything behind a login, and anything the index has not seen.