Hacker News Scraper — firehose mode, no API key

Read Hacker News's own official public API — no key, no login, no browser. Pull a curated feed snapshot (top, new, best, ask, show, job), or turn on firehose mode and get every new story, job, poll or comment since your last run, tracked by permanent sequential item id — not a time-window watermark, not a search-result diff.

$0.40 per 1,000 rows. No start fee.

---

What you get

One row per story, job or poll:

id, type, title, url, domain	the item and where it points
by, score, descendants	author, points, comment count
createdUtc, createdAtIso	exact publication time
hnUrl	direct link to the HN discussion thread

Turn on comments (feed mode) or ask for comment items (firehose mode) and each becomes its own row with text, by, parentId, storyId and depth — so a thread rebuilds from the flat table with one self-join, the same contract used by every other comment-tree Actor in this collection.

Firehose mode: the part no competing HN Actor on the Store advertises

Checked the three top-ranked competing Hacker News Actors on 2026-09-20: all three snapshot a feed or run a search. None mention a since-last-run watermark. This Actor's mode: "changes" does exactly that, and it's structurally simpler than the same feature on Reddit or LinkedIn: every Hacker News item has a permanent, strictly increasing integer id, and the API's own /v0/maxitem.json hands you the current ceiling directly. "Everything new since my last run" is therefore just the integer range between your last mark and that ceiling — no sort order to get wrong, no timestamp clock-skew, no pagination edge case.

• First run never bills your entire history. It seeds from (current max id − firstRunLookback), 500 ids back by default — a deliberate, bounded starting point, not the whole site's archive.
• A capped run never skips a gap. If maxIdsPerRun stops a run partway, the mark advances only through the longest unbroken run of ids actually processed — even under concurrent fetching, where results can complete out of order. The next run picks up exactly where the last one actually finished, never later.
• Comments carry no public score field, so minScore only ever filters stories, jobs and polls; a comment is never dropped by it.

What it needs, and what it never does

Nothing. This Actor works with no proxy and no API key at all — Hacker News's Firebase-backed API answered every request cold, and ten rapid sequential requests hit zero throttling during testing. There is no login, no session, no scraped personal data beyond what any HN user already publishes under their own public username.

Input at a glance

Input	Default	Notes
mode	list	list (a feed snapshot) or changes (firehose)
listType	top	top, new, best, ask, show, job — mode list only
itemTypes	["story"]	which types to keep in firehose mode
includeComments	false	mode list only; one request per comment, no batch endpoint exists
maxCommentsPerStory / commentDepth	50 / 5	comment-tree caps
minScore	0	stories/jobs/polls only
maxItems	200	cost cap — billed once per row saved
maxIdsPerRun	1000	mode changes only; safety cap per run, never causes a skipped id
firstRunLookback	500	mode changes only; how far back a new source's first run reaches

Billing

Pay per event, charged once per row saved, no start fee:

Event	Price	When
story-scraped	$0.0004	one story/job/poll row
comment-scraped	$0.0004	one comment row

Set maxItems as your cost cap.