Hacker News Search Scraper — Stories, Comments, Show HN & Ask HN

Search the whole Hacker News archive through its Algolia index: stories and comments with title, link, author, points, comment counts and full comment text, filtered by type, author, score and date. Keyless. No login, no browser, no proxy needed.

Input

Every field is optional, but give at least one of a search term, an item type or an author.

Field	Type	Default	Description
searchQueries	array	—	One search per entry
itemTypes	array	all	story, comment, ask-hn, show-hn, poll, front-page
author	string	—	Only items by this HN username
minPoints / minComments	integer	—	Score and discussion thresholds
createdAfterTimestamp / createdBeforeTimestamp	integer	—	Unix seconds
sortByDate	boolean	false	Newest first instead of most relevant
maxItemsPerQuery	integer	200	Capped by the index at ~1,000

Searching by filters alone works: leave searchQueries empty and ask for every Show HN above 100 points.

Output

One row shape for both stories and comments.

{
  "_input": "rust",
  "_source": "S1-hn-algolia",
  "_scrapedAt": "2026-09-09T14:02:31Z",

  "objectId": "22238335",
  "type": "story",                 // story | comment | ask_hn | show_hn | poll
  "hnUrl": "https://news.ycombinator.com/item?id=22238335",

  "title": "Why Discord is switching from Go to Rust",
  "isOwnTitle": true,              // false on a comment — see below
  "url": "https://blog.discordapp.com/why-discord-is-switching-from-go-to-rust…",
  "text": null,                    // comment body, plain text

  "author": "Sikul",
  "authorUrl": "https://news.ycombinator.com/user?id=Sikul",

  "points": 1582,
  "numComments": 642,
  "childCount": 40,

  "storyId": 22238335, "parentId": null,

  "createdAt": "2020-02-04T17:30:40Z",
  "createdAtTimestamp": 1580837440,
  "updatedAt": "2023-09-06T20:31:44Z",
  "tags": ["story", "author_Sikul", "story_22238335"]
}

Three things worth knowing

On a comment, title is the parent story's — isOwnTitle tells you which. Comments come back with story_title and story_url rather than their own, so a naive merge silently attributes a story's headline to a comment. Both are merged into title/url for a single usable row shape, and isOwnTitle is false whenever the title came from the parent. points and numComments are empty on comments because HN does not score them.

Only about 1,000 results per query are reachable. nbHits reports the true match count — 60,019 for one test — but nbPages × hitsPerPage always works out to roughly 1,000 whatever page size is requested. Paging past the end returns an empty hits list with nbPages: 0, which is honest as these things go. Slice by createdAfterTimestamp / createdBeforeTimestamp to walk a large query in windows.

Item type comes from _tags, not from which fields are missing. A comment can be spotted by its lack of a title, but that is fragile, and it cannot distinguish an Ask HN or Show HN post from a plain story. _tags states the kind outright, so that is what type is derived from.

Errors

_error	Meaning
no_results	The search ran and matched nothing
unexpected_shape	A 200 without a hits list, or a hit without an objectID
blocked	Every TLS profile was refused
network_error	The ladder never reached the server

If every search fails, the run itself fails rather than reporting success over an empty dataset.