Hacker News Search — stories, comments, live lists (Algolia + official API)

Search all of Hacker News since 2006 — stories, comments, Ask HN, Show HN, Launch HN, polls, job posts — by keyword with points / comments / date / author / domain filters, or pull the live top / new / best / ask / show / jobs lists exactly as the HN pages rank them. Optionally attach each story's discussion as a nested comment tree and the author's profile (karma, account age). Built for AI agents, brand monitors and researchers: clean JSON with plain-text bodies. No API key, no proxy, no browser.

Use cases

• Brand & competitor monitoring — every story and comment mentioning your product (or a competitor) in the last N days, with points and discussion threads, on a schedule.
• Launch research — pull all Show HN / Launch HN posts on a topic, sorted by points, to see what resonated and how people reacted (comments included).
• Content & PR intelligence — which stories from a domain (arxiv.org, github.com, your blog) hit HN, when, and how big they got.
• Lead / expert discovery — find people who comment knowledgeably about a technology; include their profile (karma, account age, bio) to qualify them.
• Trend tracking for agents — a daily snapshot of the live front page / best list with a points floor, deduped by URL, trimmed to the fields you need.
• Hiring signals — YC job posts (kind: job, searchable history, or live jobs) filtered by stack keywords.

Input

Field	Type	Default	Notes / allowed values
query	string	—	Keywords. Algolia syntax: "exact phrase", -excluded. Empty = list by date with the filters only.
queries	string[]	[]	Extra queries processed in the same run (batch monitoring). Each item carries its query. maxItems is the total.
kind	enum	story	Search (Algolia): story, comment, ask_hn, show_hn, launch_hn, poll, job, front_page, newest, any. Live lists (Firebase): top, new, best, ask, show, jobs. See reference.
sort	enum	relevance	relevance (text match → points → comments; falls back to date when no query), date (newest first), points, comments. Points/comments are sorted client-side: the actor fetches the whole reachable pool (up to 1,000 hits, 5 requests) and returns the top maxItems.
searchIn	enum	all	all, title, url, text, title_text, author → Algolia restrictSearchableAttributes. all also matches story text, URLs and (for kind: comment/any) comment text — right for mention monitoring, noisy for topic lists: use title for "stories about X".
typoTolerance	bool	false	Let Algolia match misspellings. Off by default because it turns short terms into noise ("LLM" → "Limbo", "Limits", "GLM"; "Kafka" → "Kakapos"); turn on to also catch misspelled mentions.
minPoints	int	0	Algolia numeric filter points>=. Ignored for comments (they have no points).
minComments	int	0	Algolia numeric filter num_comments>=.
sinceDays	int	0	Last N days; 0 = all time. Overridden by dateFrom.
dateFrom / dateTo	date	—	YYYY-MM-DD or ISO 8601 UTC, inclusive.
author	string	—	Bare username (pg) → tag author_pg.
storyId	string	—	Restrict to one thread → tag story_<id> (use with kind: comment).
domain	string	—	Keep items whose URL is on this domain or a subdomain (client-side). With an empty query it becomes the URL query, so "all submissions of a site" is one call.
maxItems	int	50	1–5000, total across queries. Algolia serves at most 1000 hits per query; live lists hold 200–500 ids. Legacy alias limit still works.
includeComments	bool	false	Attach nested comments per story (1 extra request per story that has comments).
maxCommentsPerStory	int	20	1–500, breadth-first (top-level replies first).
commentsDepth	int	2	1–10 reply levels.
includeAuthorProfile	bool	false	Attach authorProfile (karma, createdAt, about, submittedCount) — 1 request per distinct author, cached in the run.
textMaxLength	int	2000	100–50000 chars for text and comment texts.
dedupeByUrl	bool	false	One item per external URL (scheme/www/trailing slash ignored).
fields	string[]	[]	Whitelist of output fields (id always kept).

Filters are ANDed. Live Firebase kinds ignore query, author, storyId, dates and searchIn (they are not searchable) and apply minPoints / minComments / domain client-side — the run logs a warning when an input is ignored.

Reference

kind values

kind	Source	What you get	Filters
story	Algolia tag story	Stories (link + text posts), all time	all
comment	Algolia tag comment	Comments with parent story context	all except minPoints/minComments
ask_hn / show_hn / launch_hn	Algolia tags	Ask HN / Show HN / Launch HN (YC launches) posts	all
poll	Algolia tag poll	Polls (options are type pollopt)	all
job	Algolia tag job	YC job-board posts, full history	all
front_page	Algolia tag front_page	~30 stories currently on the front page	all
newest	Algolia search_by_date, tag story	Alias for story + sort: date	all
any	Algolia, no type tag	Stories, comments, polls, jobs together	all
top	Firebase topstories	Live HN front page ranking, up to 500	minPoints, minComments, domain
new	Firebase newstories	Live newest, up to 500	same
best	Firebase beststories	Live "best" ranking, up to 200	same
ask / show / jobs	Firebase askstories / showstories / jobstories	Live Ask HN / Show HN / jobs lists	same

Algolia tags used under the hood (tags output field)

story, comment, poll, pollopt, job, show_hn, ask_hn, launch_hn, front_page, author_<username>, story_<id>. Tags are ANDed with commas; the actor builds them from kind, author and storyId.

Algolia numeric filters

points, num_comments, created_at_i (unix seconds) with <, <=, =, >=, >; built from minPoints, minComments, sinceDays / dateFrom / dateTo.

searchIn → restrictSearchableAttributes

searchIn	attributes
all	title, url, story_text, comment_text, author (Algolia default)
title	title
url	url
text	comment_text, story_text
title_text	title, story_text, comment_text
author	author

Topic vs mention searches. The default searchIn: "all" finds every item where the word appears anywhere — including self-post text and URLs — which is what brand monitoring needs (kind: "any" adds comments). For "the best stories about X" lists, set searchIn: "title": e.g. LLM stories of one week sorted by points returned 191 hits with all (among them a post about migrating prompts to Ollama and a "Pelican-bicycle alternatives" Show HN that mention LLMs only in their text) vs 140 with title, all with LLM in the headline.

Examples

Brand monitoring, last 30 days, stories with traction plus their discussion

{ "queries": ["apify", "crawlee"], "kind": "story", "sinceDays": 30, "minPoints": 5, "sort": "points", "includeComments": true, "maxCommentsPerStory": 10, "commentsDepth": 2, "maxItems": 100 }

What did people say about a topic? (comments, newest first)

{ "query": "\"vector database\"", "kind": "comment", "sinceDays": 90, "sort": "date", "maxItems": 200 }

All Show HN launches about a category, ranked by points, with founder profiles

{ "query": "MCP server", "kind": "show_hn", "dateFrom": "2026-01-01", "sort": "points", "includeAuthorProfile": true, "maxItems": 50 }

Every HN hit from a domain this quarter (PR report)

{ "domain": "arxiv.org", "dateFrom": "2026-07-01", "dateTo": "2026-09-30", "minPoints": 50, "sort": "points", "dedupeByUrl": true, "maxItems": 300 }

Daily snapshot of the live top list, compact fields

{ "kind": "top", "minPoints": 100, "maxItems": 30, "fields": ["title", "url", "domain", "points", "numComments", "createdAt", "hnUrl"] }

Whole thread of one story as flat comments (searchable)

{ "kind": "comment", "storyId": "4058874", "query": "scraping", "maxItems": 100 }

Output

One dataset item per story / comment. Example (trimmed):

{
  "id": "46513489",
  "type": "story",
  "kind": "story",
  "query": "apify",
  "title": "Show HN: mcpc – Universal command-line client for Model Context Protocol",
  "url": "https://github.com/apify/mcpc",
  "domain": "github.com",
  "hnUrl": "https://news.ycombinator.com/item?id=46513489",
  "points": 50,
  "numComments": 5,
  "author": "jancurn",
  "createdAt": "2026-01-06T14:03:11Z",
  "text": null,
  "storyId": "46513489",
  "storyTitle": null,
  "storyUrl": null,
  "parentId": null,
  "tags": ["story", "author_jancurn", "story_46513489", "show_hn"],
  "comments": [
    { "id": "46513489", "author": "stinny", "createdAt": "2026-01-06T15:29:16.000Z", "text": "Pretty sweet tool…", "hnUrl": "https://news.ycombinator.com/item?id=46513489",
      "children": [ { "id": "46514078", "author": "jancurn", "text": "Thank you! …", "children": [] } ] }
  ],
  "authorProfile": { "username": "jancurn", "karma": 748, "createdAt": "2011-10-31T16:07:06.000Z", "about": "Jan Čurn — Founder & CEO @ Apify …", "submittedCount": 121, "hnUrl": "https://news.ycombinator.com/user?id=jancurn" },
  "source": "algolia",
  "sourceUrl": "https://hn.algolia.com/api/v1/search?query=apify&tags=story&numericFilters=points%3E%3D5%2Ccreated_at_i%3E%3D…&hitsPerPage=100&page=0",
  "fetchedAt": "2026-09-12T23:32:10.512Z"
}

Field	Type	Meaning
id	string	HN item id (stable; hnUrl = news.ycombinator.com/item?id=<id>)
type	string	story, comment, poll, pollopt, job
kind, query	string	Input that produced the item (query null for live lists)
title, url, domain	string|null	Submission title, external URL, hostname without www.
points, numComments	int|null	Score and comment count (comments: points null)
author, createdAt	string	Username, ISO 8601 UTC
text	string|null	Story/comment body as plain text: HTML stripped, entities decoded, paragraphs kept, links as label (href), truncated to textMaxLength
storyId, storyTitle, storyUrl, parentId	string|null	Thread context (for stories storyId = id; the rest only on comments)
tags	string[]	Algolia tags (see reference)
comments	array	Only with includeComments: nested {id, author, createdAt, text, hnUrl, children[]}
authorProfile	object|null	Only with includeAuthorProfile: {username, karma, createdAt, about, submittedCount, hnUrl}
source, sourceUrl	string	algolia or firebase and the exact API request
fetchedAt	string	Fetch timestamp

Dataset views: Overview (stories) and Comments.

Use it from code / agents

curl (Apify API, run synchronously and get the dataset):

curl -X POST "https://api.apify.com/v2/acts/yadroo~hackernews-search/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H 'content-type: application/json' \
  -d '{"query":"apify","kind":"story","sinceDays":30,"minPoints":5,"maxItems":50}'

JavaScript (apify-client):

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('yadroo/hackernews-search').call({ query: 'apify', kind: 'story', sinceDays: 30, minPoints: 5, includeComments: true });
const { items } = await client.dataset(run.defaultDatasetId).listItems();

Python (apify-client):

from apify_client import ApifyClient
client = ApifyClient(token)
run = client.actor("yadroo/hackernews-search").call(run_input={"kind": "top", "minPoints": 100, "maxItems": 30})
items = client.dataset(run["defaultDatasetId"]).list_items().items

MCP: add https://mcp.apify.com to your agent (Claude, Cursor, custom) and call the tool yadroo/hackernews-search with the same JSON input.

Pricing

Pay per event: $0.001 per run start + $0.0005 per item (story or comment; nested comments and authorProfile are part of the item, not billed separately).

Typical runs: a 30-day brand check with 40 hits ≈ $0.021; a 500-item topic export ≈ $0.251; a daily top-30 snapshot ≈ $0.016.

Limits & FAQ

• Rate limits. HN Algolia allows ~10,000 requests/hour per IP; the actor pages 200 hits per request with a short pause and retries 429/5xx with backoff, so normal runs never get close. The Firebase API has no published limit; item fetches run 8 in parallel.
• 1000-hit cap. Algolia refuses to page past 1000 hits for one query. The actor caps per query and logs a warning with the true match count — split by dateFrom/dateTo windows or several queries to go deeper.
• Freshness. Algolia indexes new items within minutes; points/comment counts are updated periodically, so very fresh stories may show lower numbers than the live site. Live top/new/best/ask/show/jobs kinds read the official API in real time.
• Sorting by points/comments is done client-side: up to 1,000 matching hits are fetched per query and the top maxItems returned. Algolia never serves more than 1,000 hits, so for a global "top of all time" on a broad query narrow with minPoints or a date range.
• Search precision. Matching is word-based with prefix and plural matching (LLM finds "LLMs"); typo matching is off unless typoTolerance: true. searchIn: "all" (default) matches text and URLs too — see the note under Reference for topic lists.
• Comments and points. HN does not expose comment scores; points is null for comments and minPoints is ignored for kind: comment.
• Errors. Invalid input (unknown kind/sort, bad dates, dateFrom after dateTo) fails immediately with a clear message before any request. Per-item enrichment failures (comments, profiles) are logged and leave comments: [] / authorProfile: null; if the source is unreachable and nothing was collected, the run fails instead of looping.
• Deleted / dead items are skipped; deleted comments are dropped from comments.
• Roadmap. Poll options (pollopt) inline with polls; user-submission listing (author + empty query already covers most of it).

Sources: HN Search API by Algolia, official Hacker News API. Content belongs to its authors / Y Combinator.

---

Made by Yadroo — more data actors for agents: stackexchange-search · github-repo-intel · google-news-search · rss-to-json · crypto-news