Questions and answers

Getting started

What works before you connect anything, what each connection adds, and how the agent gets in.

What works if I add an app and connect nothing at all?

Most of it. Adding an app seeds a starter keyword set from the app's own listing, and from there ASOHawk collects ranks for those keywords and the top results behind each one, computes difficulty, a popularity estimate and a visibility score, follows the store metadata of your app and of every competitor you track, reads reviews, and records your own apps' positions in the category charts. All of that runs on public App Store data and needs no account anywhere else.

What does connecting App Store Connect add?

Your own numbers instead of public ones: downloads with the impressions to page views to installs funnel (Audience), sales and subscriptions (Revenue, which additionally needs your vendor number), the live reviews you can actually reply to, your in-app purchase and subscription prices, your builds and the release checklist, and the private keywords field, which no public source exposes. It also lets you import an app that has no public listing yet.

What does Google Analytics 4 add?

What happens after the install: active users, stickiness, retention by cohort, the in-app funnel and which features get used. That is the Product tab, and it answers the one question the App Store cannot.

What does Apify add?

Exact Apple Search Ads popularity for the keywords you choose, measured through your own Apify account. Without it every tracked keyword still carries the free calibrated estimate. What a measurement costs and who starts it is the popularity section below.

What does SMTP add?

It lets your agent send plain text email from your own mailbox, say a summary to a co-founder or a message to Apple. Everything sent that way is logged in the Emails section. Notifications from ASOHawk itself do not need it.

How do I connect my agent?

In Settings → API keys, create a key: read covers inspection and reports, write also lets it manage tracking and file proposals. Claude Code takes a single command; Cursor, Codex and Claude Desktop take a few lines of config. The server is remote HTTP, so there is nothing to install. The key is shown once.

What was the wizard on my first sign-in?

Four steps: it creates a key and drops it straight into a ready made prompt for your agent, points you at your first app, lists what each integration unlocks, and walks the eight sidebar sections. Skipping it costs nothing; everything it covers stays reachable from the dashboard afterwards.

What can I ask my agent?

Plain language, no syntax to learn. The agent picks the tools itself and answers with the numbers behind them. Four asks people start with:

Every tool it can call, with scope, cost and when to use it: full tool reference.

And the ones people ask once they settle in

Every ask below maps onto a flow that ships today. Under each is what the agent actually does with it.

1. "Publish build 42 for review."

   Waits until Apple has finished processing that build, attaches it to the version being prepared, and files the submission for your approval. You upload the binary yourself from Xcode, Transporter or fastlane, because Apple exposes no API for that step, and the submission can never go through on auto.

2. "Draft replies to this week's 1-star reviews."

   Reads your live App Store Connect reviews, drafts a response to each and queues them for approval. Nothing is published until it is approved, by you or by a policy you set to Auto beforehand, and App Store Connect caps a response at 5970 characters.

3. "Replace the screenshots on the English listing with the new set."

   Reserves upload slots for the files, then proposes the new set for one locale and device size at a time. Replacing screenshots that are already live counts as high risk and waits for a human even when screenshots are set to Auto.

4. "Rewrite the subtitle to lead with the offline mode."

   Files a diff of exactly that field. Six text fields are open to it: app name, subtitle, description, keywords, promotional text and what's new. You approve the diff, then it publishes.

5. "Which languages should I localize next? Then write the German listing."

   Ranks the gaps by countries that send traffic with no localized listing, by the locales competitors ship, and by what your listing already fills, then proposes the copy for the new locale. The agent writes the translation itself; App Store Connect only checks lengths and locale codes.

6. "Did the new subtitle actually help?"

   Records a hypothesis with the baseline and an observation window when the change goes out, then reports the outcome once the window closes. Ask any time and it lists the ones still open.

7. "What changed since I last checked?"

   One feed of what moved: competitors' metadata, version and price changes, your keywords entering or leaving the top 10, chart moves, and hypothesis windows that closed. What waits in Approvals and what sits on your task board are two separate reads, so say you want those as well and it will fetch them too.

8. "Import my unreleased app from App Store Connect."

   Imports it by ASC app id or bundle id with no public listing needed, and hands back the pre-launch checklist: texts, screenshots, category, age rating, price, availability. Keyword tracking and rank collection switch themselves on once the app ships.

9. "Reprice the yearly subscription one tier up."

   Reads the product ids and the price points Apple offers today, then files one proposal for that one product. Money is never approved automatically, whatever the policy says, and existing subscribers keep the price they signed up at.

10. "Email this summary to my co-founder."

    Sends it as plain text through your own SMTP connection. It sends only what you explicitly asked it to send, because an email cannot be recalled.

Prompt library

Copy one of these into the agent connected to your workspace. Replace the text in square brackets when it matters. Each prompt asks for evidence and limits, so the agent does not fill gaps with guesses or make a costly change before you say so.

A prompt is an instruction, not a permission. Your key scope, Agent permissions, approval rules and integration status still apply.

Orient me in this ASOHawk workspace. Inspect the workspace state, apps, integrations, data freshness, active tasks and permissions. Give me a short baseline: what is connected, what data is ready, what is missing, and the three highest-value next steps. Do not create or change anything.

Add this App Store app to my ASOHawk workspace: [App Store URL]. It is [my app / a competitor]. Before you add it, tell me what data ASOHawk can collect for that ownership type. Afterward, report the starter keywords and first collection status. Do not add competitors or keywords beyond the starter set unless I confirm them.

Audit the integrations needed for this goal: [goal]. Tell me exactly what App Store Connect, GA4, Apify and SMTP would each add to this workspace, what is already connected, and the next human step for anything missing. Never ask me to put a secret, private key or token into chat.

Make a keyword tracking plan for [app] in [countries]. Inspect the app, its existing tracked keywords, competitors, quota and current data first. Recommend the countries and a staged keyword set with the reason for each group. Show the plan before you track, archive or change any country.

Before we work, inspect my ASOHawk scopes and Agent permissions. Explain which actions will run, wait in Approvals, or be refused. Treat metadata, prices, screenshots, releases, email and paid ASA measurements as opt-in actions: show me the proposed effect and ask when the tool contract requires confirmation. Do not change my permissions.

Check whether ASOHawk has collected enough data for [app] and [country]. Show what is available for ranks, keywords, metadata, reviews, charts, downloads and revenue, when each was last refreshed, and the exact prerequisite for every missing layer. Do not invent a metric that has not been collected.

Refresh the rank data for [app] now if the workspace rate limit allows it. Then tell me what changed in the latest snapshot and what is still on the normal collection schedule. Do not describe the result as live App Store data until the refresh has actually finished.

Research a starting keyword set for [app] in the [country] storefront and [locale] listing. Use the current listing, visible competitors and ASOHawk data. Give high-, medium- and lower-demand candidates, explain why each fits, and separate evidence from hypotheses. Do not promise downloads or search volume. Wait for my confirmation before tracking terms.

Give me a weekly report for my whole ASOHawk portfolio. Compare this week with the previous one across ranks, visibility, keywords, reviews, competitor changes, downloads and revenue where those data layers exist. Name the evidence for each conclusion, separate missing data from a zero, and finish with the three actions worth considering. Do not create tasks or proposals yet.

I want to measure exact Apple Search Ads popularity for the most useful tracked keywords in this workspace. First find terms with proxy or missing popularity where an exact reading could change a decision and estimate a selected set without spending. Ask me for a maximum USD budget, no more than $5.00. Choose a set within it, show the exact estimate and storefronts, and ask for final confirmation. Start nothing until I explicitly confirm that exact set and amount. Afterward, report the run id and its status instead of starting a second run.

Find the highest-value keyword opportunities for [app] in [country]. Compare current tracked terms, rank gaps, competitors, difficulty, popularity provenance and the listing text. Return a ranked shortlist with the expected role of each term: defend, improve, test or avoid. Do not track terms until I confirm the shortlist.

Audit the title, subtitle and keyword field for [app] in [locale]. Explain coverage, duplication, weak terms, competitor gaps and length constraints. Give three ranked rewrite directions with their trade-offs. Draft nothing for App Store Connect and do not create a proposal until I choose one direction.

For [app] in [locale], use the current listing, keyword evidence, reviews and competitor positioning to propose the smallest metadata change likely to improve [goal]. Show the exact before-and-after text, the risk, evidence and limitations. Create an approval proposal only after I approve the draft in this conversation.

Choose the next localization opportunity for [app]. Compare the locales already filled, storefront activity, acquisition data when available and competitor coverage. Rank the top three locale gaps, explain the evidence, and draft the first locale only after I select it. A new locale must be presented as a proposal, not published silently.

Study the competitors of [app] in [country]. Tell me which changes, keywords, metadata patterns, ratings or chart moves are most relevant to us right now. Distinguish observed facts from interpretation, then give a short counter-strategy. Do not add new competitors or change tracking unless I say so.

For [app] in [locale], identify the most important review or conversion themes that the listing could address. Recommend whether to draft review replies, improve screenshots, or change copy. Show the proposed response or visual brief first. Treat publishing a review reply and replacing a live screenshot set as approval-gated actions.

What changed for [app] since [date or my last visit]? Check keyword and chart movement, competitor metadata or version changes, reviews, recorded listing changes, hypotheses, tasks and approvals. Give a dated timeline, then say which changes need a decision from me.

Explain why visibility changed for [app] in [country] over [period]. Compare ranks and keyword coverage, identify which terms carried the movement, check known listing and competitor changes, and separate correlation from a proven cause. Give the next diagnostic step before suggesting a fix.

How are downloads, proceeds, subscriptions and revenue changing for [app] over [period]? Use App Store Connect data only when it is available. Break down the answer by what the reports actually contain, call out reporting delays or missing ASC roles, and do not replace unavailable revenue with a public estimate.

Diagnose retention and the product funnel for [app] over [period]. Use GA4 data if connected to find the largest drop-off, compare cohorts and feature adoption, and say whether the likely next step is product work, acquisition work, or more evidence. If GA4 is missing, tell me exactly what cannot be concluded.

Check whether my own apps compete for the same keywords in [country]. Show overlaps, the current rank of each app, the severity hint and the business trade-off. Recommend only the overlaps worth resolving. Do not modify any listing until I choose a strategy and approve a proposal.

Estimate the downloads and revenue range for this public App Store app: [App Store URL]. Give the confidence and limitations with the range. Treat it as a directional public estimate, not the developer's actual sales data, and do not add the app to my workspace unless I ask.

Show every ASOHawk approval that needs my attention. For each, summarize the exact change, risk, unresolved confirmation and expected effect. Recommend approve, reject or request a revision, but do not apply anything until I decide.

Turn the evidence from [report or goal] into a small prioritized ASO plan for this week. Separate work the agent can do, work that needs my approval, and work that needs a human outside ASOHawk. Show the plan first, then create only the tasks I explicitly confirm.

Prepare [app] and version [version] for App Review. Inspect builds, release readiness, metadata, screenshots, price, availability and missing compliance. Give me a checklist in dependency order. Do not upload a binary, attach a build, submit for review or change the listing until the relevant human confirmation is present.

Draft a concise update about [app or portfolio] for [recipient]. Use only evidence available in ASOHawk, mark estimates and missing data, and show the full plain-text email to me first. Send it through SMTP only after I explicitly confirm the exact message and recipients.

Where the numbers come from

How ranks are taken, how long they live, and what difficulty and visibility actually measure.

Where do ranks come from, and how often are they taken?

ASOHawk searches the App Store itself, separately in each storefront, six times a day. Those runs are stored in three eight-hour slots, so a keyword gets at most three history points a day and a day-over-day comparison stays stable. Metadata is checked once a day, reviews and category charts twice.

Which countries get collected?

Only the storefronts where you actually track a keyword. Adding an app seeds starter terms only in its primary storefront; ASC locales add choices to the switcher, not copied keywords. On an empty ASC locale, the Keywords tab offers a localized research prompt to copy, then you add the selected terms there. Your own apps' reviews are the one exception: ASOHawk checks every supported public storefront, independent of ASC metadata locales, and marks each review with its storefront.

How long is history kept?

Day-level ranks for 90 days, after which they are rolled up into a weekly best, average and worst rather than deleted, so long trends survive while the daily detail goes. The stored top-20 result lists behind each keyword are kept 30 days.

What is difficulty?

A 0 to 100 read on how hard the top of that term looks right now, computed only from the current top 20 results for the term in that storefront: how many ratings the incumbents carry, weighted toward the higher positions, whether the top three titles actually contain the term and how old those apps are, and what share of the top 20 has at least a thousand ratings. It moves when the top 20 moves.

What is the visibility score?

A 0 to 100 summary of how much of your tracked demand you actually capture. Each keyword's rank becomes a weight that drops off fast with position and reaches zero past rank 100, and each keyword's popularity decides how much that weight counts. First place on a term nobody searches barely moves the score; slipping on a high-demand term shows up at once.

What do the confidence labels mean?

How much the number under them is worth. Difficulty is judged on sample size: a full top 20 is high, ten results medium, fewer than that low. Popularity is judged on provenance: your own exact ASA reading is high, the pooled figure medium, the free estimate medium at best and low when the estimate itself is small. Visibility carries no label.

Can I force a fresh reading?

Ask your agent to refresh an app and it queues a rank pass at once, at most once per app every ten minutes. It refreshes ranks only; metadata, reviews and charts stay on their own schedules. Tracking a new keyword already queues a collection for it, so there is nothing to force in that case.

Keyword popularity: the free estimate vs an exact measurement

Search popularity is the one number ASOHawk cannot derive on its own. Apple sells it, so every tracked keyword carries a free estimate, and the exact figure is there to buy when a decision turns on it.

Where does the popularity on a tracked keyword come from?

By default it is a proxy estimate ASOHawk computes for free, fitted against real Apple Search Ads readings so it sits on the same scale. Every keyword also carries its source, one of four values. proxy is that free estimate. asa is an exact reading this workspace measured itself. asa-aggregate is a pooled figure built from exact readings other workspaces paid for, published only once at least three independent workspaces have measured the same keyword, so it never exposes what any one of them did. none means the keyword carries no popularity number at all yet, not even the free estimate.

When is it worth measuring exactly?

When the exact number would change what you do next. For sorting a long list of candidates the free estimate is already enough. To pick the set worth paying for, ask your agent for tracked terms whose popularity_source is still proxy or none, then narrow it with a rank band (rank_max) or a floor on the estimate (min_popularity).

What does an exact measurement cost, and who bills it?

ASOHawk uses the Apify account connected to this workspace; it does not send you to an external Apify task. The assumed rate is about $0.023 per keyword, and a keyword is one term in one storefront, so a set spread over several countries costs no more than the same number of keywords in one. Copy the ASA snapshot prompt from Keywords: it reads Run ASA snapshots first. Auto lets the agent start a priced set under the $5 cap; Ask makes it collect a budget, show the exact estimate and ask for final confirmation. Each run has a $5 hard limit enforced by the worker and Apify. It is an estimate rather than a quote: it prices every actively tracked keyword in that set, drops the rest and tells you it did, and rounds up to the next cent so it never reads below the rate it assumes. That rate is an assumption, not a tariff ASOHawk can see, so what Apify actually charges depends on their live pricing and on the credits left on your account. The free plan includes $5 of credits a month, and small sets usually fit inside it.

Can my agent start a paid run on its own?

Yes, when Run ASA snapshots is Auto in Settings → Agent permissions: that is standing authorisation to start a priced set up to $5. With Ask, the agent must collect a budget and explicit final confirmation in the conversation instead. The worker repeats the applicable cap just before it calls Apify.

What changes after a run succeeds?

The terms that got a reading flip to popularity_source: asa, and the number you then see is Apple's own. Terms with too little demand for Apple to report come back without a reading, keep whatever source they had, and are still billed. Measuring a keyword that already reads asa buys the same value twice, so your agent will not suggest it.

Approvals and safety

What the agent may do alone, what waits for you, and where your credentials sit.

What does the agent do on its own?

Reading, and managing what ASOHawk tracks: adding your own apps, tracking and archiving keywords, and setting countries. None of that touches your App Store listing, so none of it needs an approval. Adding a competitor is the exception among those: it does not touch your listing either, but it spends your app quota and starts daily collection, so by default the agent has to confirm you asked for that app before it can track it.

What waits for me?

Anything that would change what customers see unless its write policy is Auto. The agent files a proposal with the exact diff. Ask parks it in Approvals; Auto approves it by policy and it can reach App Store Connect without a separate click. Prices, category, age rating, availability, native A/B tests and App Review submissions still wait for a human.

What are Auto, Ask and Deny?

Settings → Agent permissions sets one of the three per operation type. On a change to your listing, Auto applies the proposal without waiting, Ask parks it in Approvals, and Deny refuses the call outright. Operations with no approval queue read Ask differently, and each tool's entry in the reference says which way: tracking a keyword or attaching a build simply runs. For an exact popularity measurement, Auto is standing authorisation to spend up to $5; Ask requires a budget and final user confirmation instead. Adding a competitor reads Ask the same strict way: the agent has to confirm you asked for that app before it can track it. A setting a new workspace never touched reads as Auto, except Add competitors, which reads as Ask. Workspaces that existed before this default changed keep their previous effective Ask entries until an owner changes them. The same page carries a separate Reading list, where each data domain is Allow or Deny for the agent, with no queue in between.

What can never be automatic?

Submitting a version to App Review and starting a native A/B test, whatever the policy says. Also replacing a screenshot set that is already live, and any change to money or reach: app price, an in-app purchase or subscription price, category, age rating, and the territories the app sells in. One exception exists and the agent must ask for it explicitly: a purchasing-power price matrix where every territory already has a price and none moves more than 15%. A high-risk rewrite of metadata text is automatic only when its High-risk changes policy and every affected field are Auto.

Can the agent spend my money?

Only after your explicit consent. Exact popularity measurement is the paid action: the agent must first get a budget, show the exact selected estimate, and receive your final confirmation. The run cannot exceed that budget or $5, whichever is lower.

Where do my credentials live?

Encrypted, under envelope encryption, and never in plain text, never in a log and never displayed back: the Integrations page shows connection status only. The App Store Connect private key is read in your browser and sent once, never written to disk. Disconnecting an integration deletes its credentials immediately.

What can an API key do?

Read or write, chosen when you create it. A write-only tool is invisible to a read key rather than merely refused. The key is shown once; revoke it and any agent using it stops immediately. If the member who created a key is suspended from the workspace, that key pauses with them and resumes when they return, while keys other members made keep working.

Notifications

What can reach you without opening the dashboard.

What arrives the moment it happens?

Four alerts, each switched on separately in Settings → Notifications: a tracked keyword drops out of the top 10, a new 1-2 star review lands on one of your apps, a tracked competitor changes name or subtitle or ships an update, and one of your own apps enters, leaves or jumps in a top chart. A broken integration is mailed to the workspace's owners and admins on its own path, so that one arrives regardless.

What is in a digest?

Five sections, the ones needing action first: integrations that broke, keywords that dropped, new low reviews, competitor changes, keywords that gained. The daily and the weekly digest carry the same sections over a different window. A digest with nothing in it is not sent at all.

Email or Telegram?

One of the two, not both. Connect the bot first, then pick Telegram, and your alerts and digests arrive there instead of your inbox. Block the bot and delivery switches back to email, that message included; other Telegram failures, a timeout or a rate limit, are not caught that way. These settings are personal: every member of a workspace sets their own.

Tips

Four things that are easy to miss.

What each section of the dashboard is for

Eight sections in the sidebar. Your agent reads and writes the same data through MCP.

Overview

Your portfolio at a glance.

Apps

Your apps, their competitors and per-app data.

Keywords

Every tracked keyword, with popularity and difficulty.

Estimate

Downloads and revenue estimate for any App Store link.

Tasks

Work you delegate to your agent, and what it asks of you.

Learnings

What you and your agent have learned about the portfolio.

Emails

Every email sent through your SMTP integration.

Approvals

Store metadata changes waiting on your approval.

Five of them up close

The ones a one-line description cannot carry: what each is for, how to work it, and what to ask your agent.