Hn Developer Sentiment MCP

What developers really think about your tool.

A deterministic read on Hacker News developer sentiment for any brand, tool, or library — sentiment, mention volume, recurring themes, and feature requests across HN stories and comments, exposed as four agent-callable MCP tools. The differentiator is judgment and aggregation, not raw access: it scores, clusters, and mines public HN discussion so an AI agent can act in a single call, and returns the raw matched items alongside the aggregates so every number is auditable.

Zero setup — no API key. The Hacker News (Algolia) API is open, so this Actor boots and returns data immediately. No credentials, no approval queue, no first-run friction.

No LLM in the analysis layer — sentiment is a transparent lexicon-plus-rules scorer, so the same posts always produce the same answer.

Who is this for?

• AI agents that need a one-call read on how developers feel about a product before drafting a comparison, a launch post, or a support reply.
• DevRel and product teams tracking how a tool, framework, or release is landing with the Hacker News crowd.
• Founders and marketers validating positioning against what developers actually say in public.

Why use Hacker News Developer Sentiment?

• Zero setup — no API key, no login, no approval process. Deploy and call.
• One call, a usable answer — sentiment label, net score, volume, top themes, and representative samples in a single structured response.
• Deterministic and auditable — no model temperature, no per-call drift. Same posts plus same config always produce the same output, and the raw items ship alongside the aggregates.
• Built for agents — a Standby MCP server speaking the Streamable HTTP transport, so any MCP client can connect and call it directly.
• Cheap and fast — one HN payload per query, cached and shared across the query tools. No LLM cost in the analysis path.

How to use Hacker News Developer Sentiment

1. Deploy the Actor — run it in Standby mode on Apify. It boots with no input required and returns data immediately.
2. Connect your MCP client to the Standby endpoint (https://<your-standby-url>/mcp) using the Streamable HTTP transport.
3. Call get_sentiment_summary with a query (a brand, tool, or topic) for the headline read.
4. Drill in with search_mentions, get_feature_requests, or get_trending_discussions when you need detail. The query tools share the same cached payload per query.

The four tools

query is a brand/tool/topic string (e.g. "prisma", "tailwind"). channels, windowDays, and limit are optional and fall back to the Actor's configured defaults. channels selects HN post types — story, ask_hn, show_hn, comment — and defaults to all (stories and comments, since HN opinion lives in the comments).

get_sentiment_summary — the headline read

Aggregate sentiment, mention volume, top themes, and representative samples for a query. Input: query (required), channels?, windowDays?.

{
  "query": "prisma",
  "channelsQueried": ["story", "ask_hn", "show_hn", "comment"],
  "windowDays": 30,
  "mentionVolume": 52,
  "netSentiment": 0.18,
  "sentimentLabel": "Positive",
  "breakdown": { "positive": 29, "neutral": 16, "negative": 7 },
  "topThemes": [
    { "term": "type safety", "count": 15 },
    { "term": "migrations", "count": 12 },
    { "term": "query performance", "count": 7 }
  ],
  "topSamples": [
    {
      "title": "Prisma 6 made our migrations painless",
      "channel": "story",
      "score": 312,
      "url": "https://news.ycombinator.com/item?id=48800001",
      "createdAt": "2026-06-02T14:11:00.000Z",
      "sentiment": "Positive"
    }
  ],
  "fetchedAt": "2026-07-14T00:00:00.000Z"
}

netSentiment is the mean compound score in [-1, 1]; sentimentLabel buckets it at ±0.05. topSamples are the three highest-engagement items. topThemes are stopword- and query-term-filtered unigrams/bigrams ranked by frequency.

search_mentions — matched items with per-item sentiment

Stories and comments mentioning the query, each scored, ranked by score then recency. Input: query (required), channels?, windowDays?, limit?.

{
  "query": "prisma",
  "channelsQueried": ["story", "comment"],
  "windowDays": 30,
  "count": 2,
  "items": [
    {
      "id": "48800001",
      "title": "Prisma 6 made our migrations painless",
      "excerpt": "We migrated a 40-table schema with zero downtime and the typed client caught three bugs before...",
      "channel": "story",
      "score": 312,
      "numComments": 88,
      "url": "https://news.ycombinator.com/item?id=48800001",
      "createdAt": "2026-06-02T14:11:00.000Z",
      "sentiment": "Positive",
      "sentimentScore": 0.74
    }
  ],
  "fetchedAt": "2026-07-14T00:00:00.000Z"
}

excerpt is the item text truncated to 280 characters. sentimentScore is the per-item compound in [-1, 1]. Comments have an empty title and numComments: 0.

get_feature_requests — requests and pain points, grouped

Items that voice a feature request, wish, gap, roadmap question, or pain point about the query — grouped by cue category. Input: query (required), channels?, windowDays?, limit?.

{
  "query": "prisma",
  "channelsQueried": ["story", "ask_hn", "comment"],
  "windowDays": 30,
  "count": 3,
  "items": [
    {
      "id": "48800042",
      "title": "Ask HN: does Prisma support deep JSON filtering yet?",
      "excerpt": "Please add deep JSON path filters to the query API.",
      "cueCategory": "request",
      "channel": "ask_hn",
      "score": 54,
      "url": "https://news.ycombinator.com/item?id=48800042",
      "createdAt": "2026-06-09T18:30:00.000Z"
    }
  ],
  "byCategory": { "wish": 1, "missing": 1, "request": 1, "plans": 0, "painpoint": 0 },
  "fetchedAt": "2026-07-14T00:00:00.000Z"
}

cueCategory is one of request / wish / missing / plans / painpoint (matched by priority in that order). excerpt is the matching sentence. byCategory totals always sum to count.

get_trending_discussions — highest-engagement stories (no query)

Top Hacker News stories by engagement, no query required. Input: windowDays?, limit?.

{
  "channelsQueried": ["story"],
  "windowDays": 7,
  "count": 2,
  "items": [
    {
      "id": "48841676",
      "title": "Postgres rewritten in Rust, now passing 100% of the regression tests",
      "channel": "story",
      "score": 980,
      "numComments": 240,
      "engagement": 1220,
      "url": "https://news.ycombinator.com/item?id=48841676",
      "createdAt": "2026-07-12T09:02:00.000Z",
      "sentiment": "Neutral"
    }
  ],
  "fetchedAt": "2026-07-14T00:00:00.000Z"
}

engagement is score + numComments; items are ranked by it descending. Trending is story-level (comments excluded — they have no thread engagement).

Input

The server boots with no input required and returns data immediately — the Hacker News (Algolia) API needs no credentials. All fields below are optional.

Output

Each tool returns a single structured JSON object (shown above per tool). Responses are deterministic — the same posts plus the same config always produce the same result. Every response includes a fetchedAt timestamp reflecting when the underlying Hacker News payload was pulled; a cached read returns the same payload (and timestamp) for that TTL window.

Data fields reference

Pricing

This Actor runs in Standby mode and is billed for the compute used while warm and serving requests. Reads are lightweight — each query's HN data is fetched once, cached (default 20 minutes), and shared across the query tools, so calling several tools for the same query costs a single upstream fetch. There is no LLM in the analysis path, so no per-call model cost, and the Hacker News API is free.

Tips

• Start with get_sentiment_summary for the headline, then call a detail tool only when you need it — the query tools share the same cached payload.
• Narrow channels (e.g. ["ask_hn"]) to focus on solicitation threads, or drop comment to look only at story-level signal.
• Raise cacheTtlMinutes when batching many queries; lower it when you need fresh reads.
• Widen windowDays (up to 90) to capture a steadier baseline.

How it works