AI agent prompts for App Store keyword research, with the tool calls they trigger

1. "List my apps with localaso." Call: board_get with no app. Verify: the app name and isOwn are the ones you expect.

2. "Show my board for Streakly. Which keywords have never been checked?" Call: board_get with app: "Streakly". Verify: the agent treats rank: null as "unchecked", not "not ranking". "200+" means checked and not found.

3. "Read my listing and list the words in the name and subtitle." Call: listing_get. Verify: the agent mentions the source field and whether keywordsDerived is true.

4. "Suggest 15 keywords a habit tracker could target that are not on my board yet, then add them as ideas for the US." Calls: board_get (to see what exists), then board_add_keywords with mode: "english", countries: ["US"], section: "idea". Verify: no duplicate of an existing term, at most 100 terms per call.

5. "Add these terms localized to Germany, France and Japan." Call: board_add_keywords with mode: "localize" and countries: ["DE","FR","JP"]. Verify: phrases in the result differ per country. Localize needs a provider that works without a stored key; read localize App Store keywords with AI for the details.

6. "Check the keywords you just added." Call: keywords_check with terms set to those terms. Verify: checked plus remaining equals the number of keyword and country pairs; if remaining is above zero the agent calls again. See the keywords_check page.

7. "Which keywords have a Push verdict in the US, and what is their difficulty?" Call: board_get, then the agent filters markets for country: "US" and verdict: "Push". Verify: the numbers come from the result, not from memory.

8. "Who beats me most often, and how many ratings do they have?" Call: competitors_get. Verify: the agent reports aheadCount as keyword and country pairs. See competitors_get.

9. "Draft a keyword field from my Push and Maybe keywords, no repeats of my name or subtitle." Calls: board_get, listing_get. Verify the output with the keyword field checker: the field is limited to 100 bytes, not 100 characters.

10. "Make a weekly checklist for me from the board." Calls: board_get, competitors_get. Verify: every task points to a field in the data.

• "What is the exact search volume of this keyword?" LocalASO gives a 0 to 100 estimate derived from rating counts, not Apple's count. A good agent says so.
• "Upload the new subtitle to App Store Connect." The MCP server cannot upload anything. The agent should tell you to use the app.
• "Check 500 keywords in all 52 countries now." Each keywords_check call runs at most 60 checks and the server rate-limits. Ask for batches.

State "use only data returned by tools" in the first message. Ask it to name the tool and field behind every number. Compare the final field against Apple's limits yourself; the metadata limits tool lists them. For the research method behind the prompts, see the keyword research guide.

If you have not connected a client yet, start with Claude Code or read the MCP docs.