# minradar

> minradar mines public App Store data for app opportunities a small team can actually build:
> apps with real install scale and visible dissatisfaction, plus evidence that users already pay
> in that category. Every score is explainable, and rating volume is used as a stated proxy —
> Apple does not publish downloads.

## Use it from an agent

- MCP (HTTP transport, no install): POST JSON-RPC to https://minradar.com/mcp
  - `initialize` → server info and protocol version
  - `tools/list` → the tools below
  - `tools/call` with {"name": "...", "arguments": {...}}
- REST + OpenAPI: https://minradar.com/openapi.json  (a compact `brief` is at https://minradar.com/api/app/{app_id})
- Build prompt for one app (paste straight into a coding agent):
  https://minradar.com/api/app/{app_id}/prompt?target=codex|claude|cursor
- Human-readable page per opportunity: https://minradar.com/o/{app_id}
- Everything indexable: https://minradar.com/sitemap.xml

## Tools

- find_unmet_demand: Rank App Store opportunities in apps that already have a large user base but whose users are clearly unhappy (unmet demand). Returns a ranked table with tier, score, rating and scale. Use this first when deciding what app to build.
- validate_paying_market: Rank app categories by evidence that users already pay inside them: how many of its apps reach the store-wide Top-Grossing chart, how many charge upfront, paywall language in sampled reviews, and review depth. Call this before building to confirm a market monetises.
- get_opportunity_brief: Full evidence dossier for one app: score breakdown, plain-language diagnosis, complaint themes with verbatim user quotes, review provenance, and matching open-source accelerators. Use after find_low_hanging_fruit to inspect a specific candidate.
- generate_app_brief: Produce a ready-to-build product specification for an opportunity, formatted for a specific AI coding agent (codex, claude, or cursor). Includes the wedge, the verbatim complaints to fix, verified open-source accelerators, and a verification plan.
- search_open_source_accelerators: Find maintained, commercially reusable open-source projects for a niche, with license posture. Use to avoid rebuilding infrastructure and to check whether a dependency is safe to ship.
- explain_method: Explain the scoring formulas, weights, and the honest limitations of the data. Use when a user questions the numbers.
- market_stats: Report local data coverage (apps, reviews, chart snapshots) and the last ingest run.

## Flow

`find_unmet_demand` (or `validate_paying_market`) → `get_opportunity_brief` →
`generate_app_brief`. Ask `explain_method` when a user challenges the numbers: it returns the
formulas, weights and the dataset's limitations rather than a defence.

## Limits

- No download counts. Rating volume is a proxy and is labelled as one everywhere.
- Review themes come from public review samples, not from every review ever written.
- Read-only over HTTP: writing (watchlist, ingest) needs the local stdio server.