Applyra API and MCP examples: what to build, what to ask

Four patterns come up again and again on the Applyra API: a keyword sheet that refreshes itself, a weekly movement report, onboarding a whole app in one script, and a listing review run by an AI assistant. Each is described below with the calls behind it, followed by the prompts that get the most out of the MCP server. All of them assume a key from getting started with the API, and the routes they use are documented in full in the endpoint reference.

A keyword sheet that refreshes itself

The most common one. A scheduled job calls GET /api/v1/applications once to resolve the app id, then GET /api/v1/keywords for the rows, and writes them into a spreadsheet.

Two details save you time. Ask for the app id once and cache it, since it never changes. And treat a null rank as not ranked rather than as a gap in the data: the API errors rather than returning half a ranking, so a null is a fact, not a missing value.

A weekly movement report

GET /api/v1/keywords/{id}/ranks/history returns the daily series per keyword. Diff the first and last point of the window, sort by the delta, and post the top five gains and losses to wherever your team reads.

If the whole list moved the same way on the same day, check the algorithm change detector before concluding anything about your own work: a market-wide day looks exactly like a personal catastrophe in a report. A single number that moved against the market is worth a ranking drop diagnosis instead.

Bulk onboarding a new app

POST /api/v1/applications to track the app, then POST /api/v1/keywords in batches of up to twenty. A hundred keywords is five calls in a loop, not an afternoon of clicking.

Pair it with GET /api/v1/keywords/inspect beforehand if you want to filter candidates by difficulty before committing them, which is the API version of what the Keywords Inspector does on screen.

An assistant that reviews a listing

This one needs no code. With the MCP server connected, the prompt is the program:

Read the ASO Health of my app, tell me which axis is holding it back, propose a new title and subtitle, then score your proposal with the simulator and tell me whether it beats the current listing.

The assistant chains get_aso_health, writes the draft itself, and calls simulate_metadata to have Applyra score it. You get a number rather than an opinion, which is the whole point of letting the model write and the engine judge.

What to ask an assistant

With the MCP server connected, you do not pick the tools: the assistant does, from the question. The prompts below are the ones that earn their keep, grouped by the job they do. Each one names the tools it leads to, so you can check that the assistant fetched the answer rather than guessed it. Every tool is described in the MCP tools.

Monday triage

Which of my tracked keywords lost the most ground this week? Show each one's rank a week ago and today, worst first, and skip anything that moved less than three places.

list_keywords returns the set with today's positions, and get_keyword_rank_history gives the daily series for the ones that moved. A null rank in a series is a day outside the top 100: ask the assistant to report it as not ranked, never as missing data.

Is the climb real?

Show me the 90-day rank history of my favorite keywords and tell me which ones are climbing steadily and which ones jumped once and fell back.

list_keywords carries the favorite flag, and get_keyword_rank_history the daily positions. A steady climb is worth holding; a single jump that faded is worth checking against the algorithm change detector before you credit your listing. Ninety days of history is an Unlimited feature: the free plan keeps seven, see plans and limits.

Did my last release move anything?

When was my last release, and how did my visibility score move in the two weeks before and after it? Tell me how my ASO Health score changed with it.

get_aso_health returns the listing's audit with a timeline of past releases and their score, and get_app_score_history the daily visibility score. The first says whether the listing got better written, the second whether the app got easier to find. They do not always move together, and that difference is the interesting part.

Check a draft before you score it

Here is my new Google Play short description: "…". Check it against the store's rules, then score it against my live listing.

check_metadata counts every field the way the store does and names the policy a word would break, such as price or ranking claims. simulate_metadata then scores the draft on the engine that audits your live listing and returns the gain or loss against it. The same flow runs on screen in the Metadata Simulator.

Mine the autocomplete

What does the App Store suggest in the US when someone types "habit"? Inspect the five suggestions closest to my app and tell me which ones I already track.

run_autocomplete returns the store's own suggestions, and inspect_keyword scores each one: difficulty, traffic, the KEI verdict, the apps ranking for it today, and whether you already track it. Everything the store suggests is something people type, which is what makes it a better seed list than anything the model would invent.

Pick the next keywords

From the keywords I inspected this month, keep the ones rated Easy with a traffic score of 25 or more, and track the best three for my app.

list_keyword_inspections re-reads your past research with current scores, and track_keywords adds the keepers, with their first position measured right away. It is one of the six tools that change your account: it adds to what Applyra tracks, never to your store listing. The bar it applies is the one in choosing the keywords to target.

Size up the competition

Rank me against the competitors I track by visibility score, then tell me which one is closest and how its title and description differ from mine.

list_competitors returns each pair with both visibility scores and the competitor's store listing, and list_applications your own. The assistant reads the two listings side by side, which is the fastest first pass before turning a competitor gap into a plan.

Find an open niche

Break "habit tracker" into sub-niches on the App Store in the US and tell me which one is the most open. Check first whether I already ran this topic.

list_niche_analyses shows what you already covered, and run_niche_analysis returns the sub-niches with their keywords, their opportunity score, their intent and an app concept. A fresh analysis takes a few minutes, so let the assistant check the history first. The same analysis on screen is Niche Analysis.

Watch a chart you are not in yet

Which apps climbed or entered the top free Productivity chart on the US App Store today, and what do the new entries have in common in their titles?

list_top_chart_categories resolves the category, and top_charts returns the ranking with each app's movement since yesterday and a flag on the new entries. See Top Charts for the same data on screen.

Make it show its work

One line at the start of a session keeps every answer honest:

For every number you give me, name the tool you called and the date of the data. If you did not call a tool, say the number is your estimate.

An assistant with tools still guesses when a question does not map to one. This makes the guesses visible. The prompts above, and the places where an agent gets ASO wrong, are covered at length in the agentic ASO prompt playbook.

Where to look when a call fails

Every error carries a typed code, listed in authentication, rate limits and error codes. The two you will meet first are RATE_LIMITED, which is safe to retry after the delay in the header, and QUOTA_EXCEEDED, which is not: it waits for your monthly reset.