App Store keyword research with Claude Code and LocalASO

Install and activate the app first (see the MCP server docs). Then run this in a terminal:

claude mcp add localaso -e ELECTRON_RUN_AS_NODE=1 -- \
  /Applications/LocalASO.app/Contents/MacOS/LocalASO \
  /Applications/LocalASO.app/Contents/Resources/app.asar/out/main/mcp.js

Claude Code's documentation explains that the double dash separates Claude's own options (such as --env and --scope) from the command that starts the server, and that everything after it is passed through untouched. If you ever write the flags in front of the name, put another option such as --transport stdio between --env and the server name, because Claude Code otherwise reads the name as another KEY=value pair.

Run claude mcp get localaso to see the server status, or type /mcp inside a Claude Code session. claude mcp list shows every configured server and claude mcp remove localaso removes it. You should see the five tools: board_get, board_add_keywords, keywords_check, competitors_get and listing_get.

By default Claude Code stores a server for the current project in ~/.claude.json. To commit the setup for a team, put it in .mcp.json at the project root:

{
  "mcpServers": {
    "localaso": {
      "command": "/Applications/LocalASO.app/Contents/MacOS/LocalASO",
      "args": ["/Applications/LocalASO.app/Contents/Resources/app.asar/out/main/mcp.js"],
      "env": { "ELECTRON_RUN_AS_NODE": "1" }
    }
  }
}

Claude Code asks for approval before using project-scoped servers from .mcp.json in interactive sessions. Each teammate still needs their own LocalASO license, because the server reads the license on that Mac.

These are example prompts, not recorded output. The tool calls are the ones Claude should make, based on each tool's schema.

1. Prompt: "Use localaso to list my apps." Expected call: board_get with no app. Result shape: an apps array with id, trackId, name, isOwn and mainCountry.
2. Prompt: "Show the board for Streakly and tell me which keywords have no rank yet." Expected call: board_get with app: "Streakly". Rows with rank: null have not been checked.
3. Prompt: "Add these 15 habit-tracking keywords as ideas for Germany and France, localized, then check them." Expected calls: board_add_keywords with mode: "localize" and countries: ["DE","FR"], then keywords_check.
4. Prompt: "Which of them are worth a place in the keyword field?" Claude reads the verdict, volume and difficulty fields from the check result.

Localize mode needs an AI provider that works without a stored key; the default Claude Code CLI provider does. See AI providers.

• keywords_check runs at most 60 checks per call and returns remaining; Claude must call it again for the rest. Details are on the keywords_check page.
• The agent works from estimates. Keep the final say, and verify the characters and bytes of your field with the keyword field checker.
• For ready-made prompts see AI agent prompts for App Store keywords; for the concept, what an ASO MCP server is.

Download the app on the download page, activate it, run the command above and start with the listing prompt.