tools / audit

ASO research bundle

The whole research pass — listing, word pool, scored keyword candidates — in one call.

worked example

Audit HabitKit's US listing with up to 8 discovered candidates.

{
  "name": "audit",
  "arguments": {
    "appId": 6443918070,
    "countries": [
      "US"
    ],
    "limit": 8
  }
}

{
  "appId": 6443918070,
  "storefronts": [
    {
      "country": "US",
      "app": { "trackName": "Habit Tracker - HabitKit", "subtitle": "Streaks & Accountability", … },
      "pool": [
        { "word": "habit", "field": "title" }, { "word": "tracker", "field": "title" },
        { "word": "habitkit", "field": "title" }, { "word": "streaks", "field": "subtitle" },
        { "word": "accountability", "field": "subtitle" }
      ],
      "seeds": ["habit", "tracker", "habitkit", "streaks", "accountability"],
      "candidates": [
        { "keyword": "habit tracker", "popularity": 82, "label": "Very high",
          "competition": 189, "medianTopRatings": 14002, "rank": 2 },
        { "keyword": "accountability partner", "popularity": 65, "label": "High",
          "competition": 186, "medianTopRatings": 439, "rank": 88 },
        { "keyword": "streaks", "popularity": 49, "label": "Medium",
          "competition": 188, "medianTopRatings": 12238, "rank": 3 },
        … 5 more candidates, including autocomplete noise like "the habit burger grill" …
      ]
    }
  ]
}

real response, captured 2026-07-29 — trimmed only where marked with …

• accountability partner is the finding: demand 65, incumbents with a median of just 439 ratings, and the app already at rank 88 — winnable ground the current metadata barely serves.
• Discovery is honest about its source: autocomplete also surfaces the habit burger grill (popularity 83) — an agent reading medianTopRatings: 636905 and the restaurant genre discards it in one glance.
• Expect roughly seeds + 4 × candidates seconds per storefront: every candidate is a real, throttled search against Apple. Progress notifications stream when the client requests them.

when to use it

• The first call on any new app — yours or a competitor's. It replaces the manual loop of lookup → suggest → volume → rank per candidate.
• Rebaselining after a metadata ship: rerun and watch which candidates' ranks moved.
• Feeding the drafting step: candidate popularity values become coverage phrase weights ({phrase, value}) so the 30/30/100 characters chase measured demand.
• The packaged optimize_metadata prompt on the server orchestrates this end to end: audit → draft → regression-check → verify → track.

description — as served to agents

input

output

pairs with