MCP and AI agents
What is an ASO MCP server, and what can an AI agent do with one?
An ASO MCP server is a program that exposes App Store optimization data and actions as typed tools, so an AI agent such as Claude Code can call them instead of you copying keyword tables into a chat window. MCP stands for Model Context Protocol, an open standard for connecting agents to tools.
Updated
What the server does in LocalASO
LocalASO ships a local MCP server over stdio. It opens the same SQLite database as the desktop app, so the agent sees the same boards you see. Nothing leaves your Mac except the public Apple search requests that rank checks make. The full setup is in the MCP server docs.
The server exposes five tools:
| Tool | What it does |
|---|---|
board_get | Lists your apps, or returns every keyword on one app's board with rank, volume, difficulty and verdict per country |
board_add_keywords | Adds up to 100 terms to a board, as written or localized to each country |
keywords_check | Runs rank, volume and difficulty checks and waits for them |
competitors_get | Returns the apps that rank ahead of yours on your keywords |
listing_get | Returns the saved name, subtitle, keywords, promo text and description of the main locale |
Each has its own page: board_get, board_add_keywords, keywords_check, competitors_get and listing_get.
The loop an agent follows
Most useful sessions follow the same cycle. The agent reads the board with board_get and the listing with listing_get, proposes keywords, adds them with board_add_keywords, then calls keywords_check so the new rows get ranks. Finally it reads the board again and competitors_get to see who is ahead. You decide what goes into the real listing. The example prompts page shows each step as a prompt.
What it cannot do
- It cannot upload to App Store Connect. Drafting and uploading stay in the app.
- Every tool refuses to run when the app has no valid license on that Mac and returns a message telling you to open LocalASO.
- Keys the app stores are encrypted for the app only, so the server cannot read them. Localization through the agent needs an AI provider that works without a stored key; for a local-model setup see Ollama for ASO.
- Volume and difficulty are 0 to 100 estimates derived from rating counts, not Apple search counts. See keyword difficulty.
- The server has its own rate limiter, 15 requests per minute by default, so a big check in the app and the server at once can slow both.
Why use one instead of a chat window
A chat window only knows what you paste. A tool-connected agent reads current ranks, can verify its own suggestions and can work through 50 keywords across several countries without you touching a spreadsheet. The trade-off is that the agent acts on estimates, so read its conclusions with the same care you would give any keyword score. For the research method itself, start with the App Store keyword research guide.
Next step
Pick your client: Claude Code, Claude Desktop, Cursor, Codex CLI or Gemini CLI. Download the app from the download page; the license is a one-time $29.
Questions
- Does an ASO MCP server send my keywords to the cloud?
- Not in LocalASO. The server runs on your Mac and opens the same local database as the app. The only outside traffic is the public Apple search requests that rank checks make.
- Which AI clients can use it?
- Any client that can start a stdio MCP server. The generic client page lists the three values every client needs.
- Can the agent publish my metadata?
- No. The MCP server cannot upload anything to App Store Connect. It reads and fills the keyword board and reads the saved listing.
Do this for every country, in minutes