App Reviews & Sentiment — Google Play + App Store

Monitor any app like a product analyst: metadata, recent reviews and a structured AI summary (recurring complaints, feature requests, sentiment) — from one call.

Data comes from Apple's own public endpoints (Lookup, Search, Reviews RSS): no login, no HTML scraping, no anti-bot, no proxy needed.

Google Play support arrives in phase 2. Every output row carries platformStatus so an agent can filter safely: stable (App Store — live today) vs beta (Google Play — phase 2).

---

What it's for

• Competitive / product intelligence on any app (yours or a rival's)
• Review monitoring & support benchmarking
• Quick sentiment read for a product decision or due diligence (VC / M&A)
• ASO research and pain-point discovery for marketing

Why an agent likes it: you ask for one app, you get 3 structured rows — metadata, reviews, and a ready-to-use summary — without pulling 5,000 raw reviews into your context.

Two modes, one tool

Direct — you know the app id:

{
  "mode": "direct",
  "apps": [{ "platform": "app_store", "appId": "324684580", "country": "fr" }],
  "maxReviewsPerApp": 200,
  "includeSummary": true
}

Search — discover apps by keyword, then fetch their reviews:

{
  "mode": "search",
  "searchTerm": "CRM",
  "country": "fr",
  "maxAppsFromSearch": 3,
  "maxReviewsPerApp": 100
}

Output (one row per kind)

kind	content
app	metadata: name, developer, category, rating & ratingCount, price, version, last update, icon + run diagnostics (status, reviewsFetched, platformStatus)
review	one row per review: author, rating, date, title, text, appVersion, helpfulCount
summary	opt-in: sentimentScore (-1..1), topComplaints[], topFeatureRequests[], topPraises[] — each theme with mention count and a real quote

{
  "kind": "app", "platform": "app_store", "platformStatus": "stable",
  "appId": "324684580", "name": "Spotify : Musique et podcasts",
  "rating": 4.3, "ratingCount": 912000, "reviewsFetched": 200, "status": "ok"
}

The AI summary uses one Claude Haiku call on up to 100 reviews (cost-controlled by design). Themes are grounded in the reviews — a theme needs at least 2 mentions to appear, and every count is verified against the sample.

Input reference

field	meaning
mode	direct (known ids) or search (keyword discovery first)
apps[]	{platform, appId, country, language}
searchTerm / searchPlatforms	keyword discovery (search mode)
maxAppsFromSearch	top N results per store (1-20, default 5)
maxReviewsPerApp	0 = metadata only; App Store RSS caps at ~500 reviews/app
reviewsSort	newest (default) · most_relevant · rating_low / rating_high
reviewsSince	only reviews after this ISO date
includeSummary	opt-in AI summary (charged separately)
summaryFocus	all · bugs_and_complaints · feature_requests · sentiment_only

Pricing (pay-per-event)

Event	Price
app-metadata	$0.001 / app
app-review	$0.00015 / review ($0.15 per 1,000)
app-summary	$0.02 / app (1 LLM call ≤100 reviews)

Example: 200 reviews + summary on one app ≈ $0.05. No subscription, pay per use.

Schedule it weekly

Review sentiment drifts. Point a weekly schedule at your competitor set (or your own app) to watch complaints and feature requests trend over time — same inputs, recurring run, delta you can actually act on. Agents can also subscribe to the run dataset via webhook for a continuous review-monitoring pipeline.

Reliability & honesty

• Runs never crash on a bad app id: failures come back as status: not_found | error rows and the run continues — one bad app never fails a batch.
• Apple's public RSS does not expose developer replies → developerReply: null (documented, not a bug).
• Zero setup for you: no credential, no proxy, no config — the summary key lives on the actor, buyers never touch it.

Roadmap

• ✅ Phase 1: App Store (platformStatus: stable) — direct + search + reviews + AI summary
• 🔜 Phase 2: Google Play — same schema, same summary (actor stays a single tool)
• 🔜 Rating pagination beyond the RSS window