App discovery

App Discovery

Turn a vague market question ("what's out there in sleep tracking?") into a concrete, data-backed shortlist.

Requirements

Works best with the GetAppNiche MCP server connected (search_apps, get_app_detail). Without it, fall back to reasoning from public store pages and say so explicitly.

Workflow

1. Frame the niche as filters, not just words. Restate the request as (a) a store — apple, google, or both; (b) categories to include; (c) a search phrase if the niche is narrower than a category; and (d) the thesis expressed numerically. search_apps takes the whole Explore filter set, so state the thesis directly instead of fetching broadly and filtering by eye:

   • "who makes real money here" → min_revenue: 10000, sort_by: "revenue"
   • "beatable incumbents" → min_revenue: 10000, max_reviews: 1000
   • "is this niche still being entered" → max_days_since_release: 365
   • "what's heating up" → growth_period: "30d", growth_direction: "up", min_growth_pct: 20, sort_by: "growth"
   • "abandoned apps to displace" → min_reviews: 1000, plus updated_before a year ago
   • "non-US opportunity" → languages: ["de", "ja"]

   Confirm only if genuinely ambiguous.

2. Run two or three sharp queries, not ten scattershot ones. One query per thesis (limit: 25 is a good default). Because the filters do the narrowing, a second query should test a different hypothesis — not a synonym of the first.

3. Deduplicate and rank. Merge results, then rank by whatever signal matches the user's goal. sort_by accepts downloads, revenue, growth, reviews, rating, title, released and updated, so let the query do the sorting.

4. Deepen the top candidates. For the top 5–8 apps, call get_app_detail to pull ratings, review counts, and metadata before making claims about any single app. It accepts the canonical id (apple:284882215, google:com.example.app), a bare Apple track id, an Android package name, or a store URL pasted straight from a browser.

5. Summarize the niche. Close with 2–3 sentences on the shape of the market: concentration (one winner vs. long tail), price patterns, and obvious gaps.

Output

A table — App · Store · Downloads/mo · Revenue/mo · Rating (count) · One-line angle — followed by the niche summary and a "worth a deeper look" pick with reasoning. State the filters you used, so the user can see what the shortlist is and isn't.

Guardrails

• Use revenue and download figures to rank and compare; round them in output rather than quoting to the dollar.
• Each call costs 1 credit. If a call reports an out-of-credits error, relay it: credits refresh monthly and can be topped up in Settings at app.getappniche.com.
• A filter combination that returns nothing is a finding ("nobody is making $50K/mo here"), not a failure — report it before loosening the filters, and say which constraint you relaxed.
• Don't invent apps to pad a thin niche — a short honest list is the finding.