Hacker News Monitor — Keyword Alerts for Stories & Comments

Get new Hacker News stories and comments that mention your brand, product, competitors, or any phrase. Schedule it and each run returns only mentions you haven't seen yet.

• Exact-phrase matching. "postgres" won't match "posters", and "open source" matches only that phrase, not the two words anywhere in a comment.
• Alerts, not dumps. On a schedule, you get only new mentions and pay only for those. A run with nothing new costs just the run start.
• One row per item. A comment that matches two of your keywords is returned and charged once.
• Fast and light. Uses the public HN Search API. No browser, no proxies, no login.

What you can do with it

• Get alerted when someone mentions your product or company on Hacker News, so you can join the discussion.
• Track what HN says about your competitors.
• Watch for launches, "Show HN" posts, or discussions about a technology you care about.
• Feed new mentions to Slack, email, or an AI agent for summaries and sentiment.

Input

Field	Required	Description
keywords	yes	Words or phrases, each matched as an exact phrase (case-insensitive). Up to 20 per run.
searchIn	no	both (default), stories, or comments.
sinceHours	no	Only items from the last N hours, 1–720 (30 days). Default 24.
minPoints	no	Only stories with at least this many points. Comments are not filtered.
onlyNew	no	Skip items a previous run with the same keywords already returned. Default true.

{
  "keywords": ["postgres", "supabase", "neon database"],
  "searchIn": "both",
  "sinceHours": 24
}

Monitoring: schedule this input hourly or daily. The first run returns the last 24 hours; each later run returns only what's new. One-off search: set onlyNew to false and sinceHours to the window you want.

Output

One row per story or comment, newest first per keyword:

{
  "type": "mention",
  "keyword": "postgres",
  "kind": "story",
  "id": "49840594",
  "hnUrl": "https://news.ycombinator.com/item?id=49840594",
  "title": "700 MB/s of Kafka throughput, on Postgres",
  "url": "https://rynr.dev/blog/700mbskafgres/",
  "text": null,
  "points": 4,
  "numComments": 0,
  "storyId": "49840594",
  "parentId": null,
  "createdAt": "2026-09-25T05:49:26Z"
}

For comments, kind is "comment", text is the comment as plain text, title and url are those of the story it was posted under, and points and numComments are null (HN doesn't publish comment scores). hnUrl always links to the item on Hacker News. Author usernames are not included.

Errors: free rows

{ "type": "error", "target": "x", "reason": "invalid_input", "message": "Keyword \"x\" is too short." }

reason is one of invalid_input, blocked, rate_limited, network_error. A per-keyword run summary is saved to the OUTPUT record of the run's key-value store.

Pricing

Pay per event: $2 per 1,000 mentions ($0.002 each), plus $0.002 per run start. error rows are free, and with onlyNew you never pay twice for the same mention.

Set a maximum charge per run to cap spending. When the run reaches it, it stops cleanly and says so. Mentions that didn't fit are returned on the next run.

Limits

• Up to 20 keywords per run and a look-back of up to 30 days.
• At most 1,000 stories and 1,000 comments per keyword per run (a limit of the HN Search API). For very common words, the run summary says when this cut results off; use a shorter look-back or a more specific phrase.
• Author usernames are left out by design.

Use with AI agents

The input and output are flat and predictable. The type field on every row tells mentions and errors apart, and every error has a fixed reason code. An agent can call this Actor with just keywords.