Use case
ASO MCP Server for Claude Code, Codex and More
AppMeerkat includes a local MCP server, so Claude Code, Codex, OpenCode, Cursor or Claude Desktop can read your ranks and do ASO work with your approval.
An ASO MCP server lets an AI client like Claude Code or Codex read your app’s keyword ranks and do App Store optimization work with real data instead of guesses. AppMeerkat ships one inside the desktop app: it runs only on your computer, connects to the AI you already use, and asks your approval before anything changes.
The Job
AI is good at the thinking part of ASO: grouping keywords, drafting a subtitle in 30 characters, proposing native phrases for 12 countries. It is bad at knowing facts it was never given. Ask a chatbot “what should my keyword field be?” and it will invent volumes and ranks.
The Model Context Protocol (MCP) fixes that by letting the AI call tools that return real data. An ASO MCP server gives your AI the facts: your ranks, the top 10 for each keyword, Demand (Est.) and Difficulty (Est.), competitor gaps, your current listing.
How People Do It Today
- Pasting CSVs into a chat. It works for one question, then the data is stale.
- Asking the AI to guess. Fast and confident, often wrong. Invented search volumes are the most common failure.
- Writing their own scripts against store pages. Fragile, and you maintain it.
How AppMeerkat Does It
AppMeerkat runs an MCP server on your own machine, on 127.0.0.1 only, with a token unique to your install. Requests from anywhere else are rejected.
Connect in one click. Settings › AI has a Connect button for Claude Code, Codex, OpenCode, Cursor and Claude Desktop. It shows you the exact config it will write, writes it after you confirm, and checks that it works. For Claude Code, it’s the equivalent of a single claude mcp add command, which AppMeerkat also shows so you can run it yourself.
Read tools let your AI list apps and keywords, get rank history, read the country board, get scores, list competitors and their keywords, find keyword gaps, get suggestions, read and audit your listing, and search the store live.
Write tools let it add or remove keywords, localize keywords, set tags and notes, add competitors, run brainstorms and discovery, start checks, and create listing drafts. Every write opens an approval dialog in AppMeerkat showing the client’s name and the exact change. Nothing happens until you click Apply.
Or use the panel inside the app. If you’d rather not leave AppMeerkat, the agent panel runs your installed Claude Code, Codex or OpenCode, or a local Ollama model, with only AppMeerkat’s tools enabled. No shell, no file access.
What a Session Looks Like
You, in Claude Code: “Which of my keywords dropped out of the top 10 this week in the UK, and what would you change in the subtitle?”
- The AI calls
rankings_getfor your app in the United Kingdom with a 7-day compare. - It calls
listing_getto read your current subtitle. - It calls
store_searchto check the top 10 for the dropped keywords. - It proposes two subtitles, each 30 characters or fewer, and explains the trade-off. (You can check any subtitle yourself in the free title and subtitle preview.)
- You ask it to save one. It calls
draft_create; AppMeerkat shows the diff; you click Apply.
Every number in that answer came from a tool call. The built-in instructions tell the agent never to state a number it didn’t get from one.
How It Works
- Install AppMeerkat and add your app.
- Open Settings › AI and click Connect for your client.
- Run the Ready check: AppMeerkat sends a test question and shows the result, or names the problem (not installed, not signed in, out of usage, update needed).
- Start with a prompt: Weekly Review, Find Keyword Gaps, Localize My Listing, Brainstorm a Niche, or Explain My Drops.
- Approve or decline each change.
Our guide to ASO with Claude Code walks through real prompts. For localization work through the agent, see App Store localization.
Honest Limits
- It ships with the app. The MCP server isn’t a hosted service or a separate package. AppMeerkat must be installed, and it starts in the background when your client calls it.
- You bring the AI. We don’t sell AI usage. Your client’s own limits and terms apply.
- Local models vary. With Ollama, pick a model that supports tool calling; small models can struggle with long multi-step tasks.
- The AI can still be wrong about judgment calls. That’s why every change waits for your approval.
- Drafts only. No tool submits anything to Apple or Google.
- AppMeerkat is launching soon. Get notified at launch.
Questions
Is the AppMeerkat MCP server a separate product?
No. It ships inside the AppMeerkat desktop app and runs on your computer while the app is installed. There's nothing to host and no separate download.
Which AI clients work with it?
Claude Code, Codex, OpenCode, Cursor and Claude Desktop, with one-click setup in Settings. Any client that supports local MCP servers can connect. Inside the app, the agent panel also works with Ollama.
Can the AI change my data without asking?
No. Read tools run freely. Any tool that changes data, such as adding keywords or creating a draft, opens an approval dialog in AppMeerkat with the exact change. Nothing runs until you click Apply.
Can the AI publish my listing?
No. The tools can save drafts to App Store Connect and Google Play Console. There is no tool that submits for review, and there never will be.
Does AppMeerkat see my AI login or API key?
No. AppMeerkat never reads, stores or forwards your AI provider's credentials. Your AI client runs under your own account, as it always does.
Read next
Launching soon
Let a meerkat watch your ranks.
AppMeerkat tracks every keyword, in every country, on both stores. One email at launch, with the early-bird price.