Hacker News Search: Public Story Metadata

Export public Hacker News story metadata for topic research, reading lists, and recurring news monitoring. Keep the useful links and scores without collecting author profiles or copying full articles.

What can this tool do?

Search public Hacker News story metadata through the Algolia HN Search service. Collect topic links for product research, a reading list, or a recurring industry-news review without collecting author profiles or copying article contents.

• Search 1–20 keywords or phrases in the order supplied.
• Sort each query by relevance or newest first.
• Filter by a UTC start date and minimum points.
• Deduplicate story IDs across queries within a run.
• Limit output across the entire run, rather than separately for every query.

The service is Algolia's public HN Search API. It is not Hacker News' official Firebase API or an official Hacker News search endpoint.

What data does it return?

One row per distinct matching story. Fields are consistent; unavailable values are null.

Field	Meaning
storyId	Stable Hacker News story ID, stored as a string.
title	Plain-text story title; personal email addresses are masked.
url	Public HTTP(S) link supplied by the source, or null for a text-only story. The linked page is not fetched.
discussionUrl	Hacker News discussion URL for this story.
publishedAt	Source publication date normalized to UTC, or null if unusable.
points	Source score, or null.
commentCount	Source comment count, not the comments themselves.
query	The first query that produced this row during the run.

Author names, usernames, user profiles, comments, and full linked articles are not output.

How to use it

Paste this JSON into the Actor input:

{
  "queries": ["web scraping"],
  "maxResults": 100,
  "sort": "relevance",
  "since": "",
  "minPoints": 0
}

Input	Default	Accepted values
queries	["web scraping"]	1–20 nonblank strings, each at most 200 characters. Duplicate strings are removed. An empty array is invalid.
maxResults	100	Integer 1–5,000, shared by all queries.
sort	"relevance"	"relevance" or "newest", applied to each query.
since	No cutoff	A real YYYY-MM-DD date, inclusive from midnight UTC. Omit it or use "" to disable the cutoff.
minPoints	0	Integer 0–1,000,000.

Earlier queries may fill the global output limit before later queries are reached. The tool makes at most 50 search-page requests per run, requests up to 100 hits per page, and waits one second between page requests. The source may expose fewer pages than requested. Results can therefore be partial even when maxResults has not been reached.

For repeated topic monitoring, use "sort": "newest" and set since to the date you last checked, keeping a small overlap. Save rows in your own sheet or database keyed by storyId, then update that row when points or comment counts change. The tool does not remember previous datasets or exclude previously exported IDs between runs.

If a search response is valid but no stories match the filters, the run succeeds with zero rows and an explanatory note. An unreadable or blocked query stops without a bypass attempt; other queries may still succeed. If no query response can be read, the run fails with a source error. Inspect output notes when only part of the input succeeded.

Pricing

Pay per result: $0.30 per 1,000 stories, plus $0.001 for a run returning at least one valid story.

The one-time actor-start event costs $0.001 and is triggered only when the first valid matching story is ready. Each returned story incurs one story event at $0.0003. Invalid input, a valid zero-match search, or a run with no valid results emits neither event.

For example, 10 returned stories incur $0.001 + 10 × $0.0003 = $0.004 in Actor event charges. A partial-success run still has only one start charge, plus charges for the stories actually returned. No result is guaranteed by a search input.

Is it legal?

The tool reads Algolia's public HN Search API, using story metadata only. It does not log in, fetch private pages, collect author profiles, or bypass access controls. Failed or blocked queries are stopped rather than retried through alternative identities or proxies.

Public access is not a blanket license for every commercial use. Review the source's applicable terms and retain source/discussion links when using or redistributing results. The official HN Firebase API is a different service.

FAQ

Does a valid empty result mean the tool is broken?

No. A keyword, date, or minimum-points filter can legitimately match nothing. A readable search with zero matching rows succeeds, emits no result or start events, and explains the outcome.

Can I export all historical Hacker News stories?

No. Query pagination is bounded by the source, the 50-request safety cap, and the global output limit. This is a topic-search tool, not a complete historical archive.

Are points and comment counts fixed?

No. They are source values observed when the page is read and can change. Fetch the same story again and upsert by storyId when tracking changes.

Does it collect comments or identify users?

No. It returns a comment count and a discussion link, not comment bodies, author names, usernames, or user-profile data.

Why can two query orders produce different output?

Queries are processed in order. Results are deduplicated globally and the first queries can consume the shared maxResults budget.

Feedback

When reporting an issue, include the input, expected behavior, and the source error or output note. Use public examples only; do not include API tokens, account details, or private data.