Agent Team

Every AppApex workspace ships with a core AI growth team plus specialist agents. The core team watches the whole growth loop; the specialists activate for screenshots, paywalls, compliance, launches, crashes, pricing, and localization.

Core Agent Team

The core team is always visible in the product: Conversion, Retention, ASO, Competitor Intel, and Review Intel. These agents create the main Inbox queue and feed Managed Autopilot with actions that can be drafted, staged, or safely executed.

Conversion

Watches: paywall funnel, trial-to-paid, price/plan mix, in-app upgrade flows. Cadence: continuous, with a full sweep every six hours and on-demand runs from the dashboard.

The Conversion agent owns everything that happens between a user seeing the paywall and a charge landing. It compares variants you’re already running in Superwall or RevenueCat, ranks them by statistical significance, and surfaces the next experiment when a winner is clear. It also catches funnel regressions - typical alerts include “trial start rate dropped 6.3% in the last 24 hours” or “quarterly plan conversion fell behind annual for the first time this quarter.”

Approvals from Conversion typically execute via Superwall (variant rollouts, paywall metadata) or RevenueCat (offering changes, intro offer config).

Retention

Watches: D1/D7/D30 retention curves, churn cohorts, win-back eligibility, subscription pause/cancel reasons. Cadence: hourly polling for newly-eligible cohorts, daily roll-up.

Retention digs into where users leak out of your funnel after they convert. It segments cohorts by acquisition source, locale, and SKU, and recommends specific interventions: extending trials for a struggling segment, firing a win-back push at hour 36 of inactivity, offering pause-instead-of-cancel for users whose stated reason is “too repetitive.”

ASO

Watches: keyword ranks, search visibility, store listing metadata, screenshot performance, locale-specific opportunities. Cadence: daily rank checks, weekly metadata audit.

The ASO agent maintains a rank table for every keyword you track (up to 25 on Starter, 100 on Growth, 250 on Studio) and watches for movements outside your normal range. When a competitor’s listing refresh costs you positions, it suggests metadata edits with the keyword targets, current vs. proposed copy, and predicted recovery timeline.

Approvals execute via App Store Connect - typically a subtitle or keyword-list edit on the next listing version. Screenshot rotations require manual handoff because they involve image assets.

Spy (Competitive intelligence)

Watches:your tracked competitors’ releases, paywall changes, screenshot rotations, pricing moves, and ad creative appearing on Apple Search Ads. Cadence: daily checks with a weekly Friday intel scan.

Spy is the agent that tells you when Calm raised their annual to $79.99 the same hour it happens. Add competitors from Rivals; the agent will fetch their public listings, paywall screenshots, and pricing on each cycle and flag any deltas worth your attention.

Spy doesn’t ship changes directly - it surfaces competitor signals in Rivals and recommends defensive or opportunistic moves for your other agents to execute.

Review Intel

Watches: App Store review sentiment, theme clusters, bug surfaces, regression spikes. Cadence: every two hours.

Review Intel reads every new review across every store and clusters them by theme. When 1-star reviews spike on a specific app version, it names the regression with an engineering ticket draft attached. When 4- and 5-star reviews repeatedly mention a feature you can put in your screenshots or subtitle, it suggests the copy.

Outputs land in Inbox as either bug surfacing recommendations (which hand off to your engineering tool of choice) or ASO assist recommendations for the ASO agent to act on.

How the agents coordinate

Each agent runs independently, but they share state through a common recommendations table. When Review Intel surfaces a theme like “users love the new sleep mode,” the ASO agent picks it up on its next sweep and may propose adding “sleep aid” to your subtitle. The Spy agent, in turn, can see when Conversion runs an experiment and contextualize the result against industry benchmarks.

Specialist agents share the same recommendation and execution system. For example, a Review Intel theme can become an ASO screenshot test, a Pricing Watcher alert can become a paywall experiment, and a Privacy Manifest audit can become a pre-submission checklist.

You can pause any individual agent from Agents → pick the agent → toggle Pause. Paused agents stop scheduling new runs but their backfill remains for trend continuity.

Run history & reasoning trace

Every run produces a full reasoning trace: which queries were issued, what data came back, which tools the agent called, and how it weighed the evidence. You can review the trace from Agents run history - useful for debugging unexpected recommendations or sharing the reasoning with a co-founder.