Claude Code for
your App Store.
Claude Code works with Storewatch two ways. Storewatch’s chat can run on the claude command and your Claude subscription, with no API key. Or Claude Code connects to Storewatch’s MCP server, so you ask about downloads, reviews, keywords and crashes from the terminal while you code.

Two ways Claude Code and Storewatch work together.
Storewatch is a Mac app that keeps your App Store Connect data on your Mac. Claude Code, Anthropic’s coding agent, can sit on either side of it.
| Storewatch uses Claude Code | Claude Code uses Storewatch | |
|---|---|---|
| What it is | Storewatch’s chat, drafts and reports run on the claude command | Claude Code connects to Storewatch’s MCP server |
| Where you type | In Storewatch, in Ask Storewatch | In your terminal, next to your code |
| What it costs | Nothing extra. It uses your Claude subscription, no API key | Nothing extra. MCP is part of Storewatch |
| Set up in | Settings → AI → Claude Code CLI | Settings → AI → Claude Code → Copy Prompt |
You can use both. The first suits “write my reply to this review”. The second suits “my crash rate went up in 4.3, what changed in the code?”, because Claude Code is already in the repo.
Your Claude plan.
No API key.
Pick Claude Code CLI in Settings → AI. The chat runs on your subscription, with the same data tools.

Set up the chat with Claude Code.
- Install Claude Code and sign in with your Claude account.
- Open Storewatch → Settings → AI and choose Claude Code CLI. The app says it “uses the claude command and your Claude subscription. No key needed.” The first-run setup offers it in one click if Storewatch finds
claudeon your Mac. - Open Ask Storewatch and ask a question. If the command isn’t found, Storewatch says it isn’t installed.
Storewatch starts claude in a login shell so a Homebrew or ~/.local/bin install is found. Each question times out after five minutes. You can choose a different AI for other jobs, such as review replies.
In the terminal.
With your data.
Turn on the MCP server and paste one prompt into Claude Code. It connects itself.

Connect Claude Code to Storewatch.
- In Storewatch, open Settings → AI. Under “Use Storewatch from other AI apps”, choose Read only or Full access. It is Off until you do.
- On the Claude Code row, click Copy Prompt. The menu next to it has “Copy Command Only”.
- Paste the prompt into Claude Code. It runs
claude mcp add --scope user storewatchwith the path to the helper inside the app. The user scope means every project sees it. - Start a new Claude Code session if it says so, then ask about your apps. It calls
list_appsfirst.
If Storewatch moves to another folder, run the prompt again. Anything beyond this, such as the three access levels, is on the MCP page, and the blog guide has more prompts.
Prompts an app developer can paste.
Claude Code adds what the App Store data can’t: your code and your git history.
| Prompt | What Claude does | Tools it calls |
|---|---|---|
| “Read my git log since the last tag, check crash rate for Calorie Tracker by week, and draft What’s New for en-US.” | Combines your commits with the crash trend and drafts the text, up to 4,000 characters | get_events, get_series (crashRate), get_listing, draft_listing_changes |
| “Run the ASO audit for the US and Germany. Fix my keyword field: no repeats from the title or subtitle, under 100 characters.” | Finds words that are missing or repeated and drafts a new comma-separated keyword field | get_aso_audit, get_keyword_ideas, draft_listing_changes |
| “Read my 1 and 2 star reviews from the last 30 days and draft a reply to each, in the reviewer’s language.” | Writes the replies as text in your terminal. It can’t send them | get_reviews |
| “Did the 4.3 release change my conversion? Add a marker for it.” | Compares equal windows before and after the day, then logs the event | compare_before_after, log_event |
Storewatch has crash counts and a crash rate by version, not stack traces. For the line that crashes, Claude Code still needs your crash logs.
What Claude can change, and what it can’t.
| Read only | Full access | |
|---|---|---|
| Read | Numbers, reviews, listing, keywords, ranks, prices, subscriptions, ASO audit, markets | The same |
| Draft in Storewatch | Nothing | Listing text and price changes, which wait for you to publish or apply |
| Change locally | Nothing | Add or remove tracked keywords, log events, dismiss insights |
| Never | Publish a listing, submit for review, send a reply, apply a price, read your App Store Connect key | The same |
Your code is a separate matter. Claude Code edits it under its own permissions, and Storewatch isn’t involved. The chat inside Storewatch is also separate: it has its own controls and doesn’t follow the setting for outside apps.
Questions, answered.
Do I need an Anthropic API key?
No. Claude Code CLI in Settings → AI uses the claude command and your Claude subscription. An API key is a different provider in the same list, paid per use.
Does Claude Code see my App Store Connect key?
No. The key stays in your Keychain and the MCP helper reads only the database Storewatch keeps on your Mac. It never contacts App Store Connect.
What’s different between the chat and MCP?
The chat is inside Storewatch and Claude Code runs behind it. MCP goes the other way: you work in Claude Code and it calls Storewatch’s tools. The data tools are the same.
Does it work with Codex and Cursor too?
Yes. Settings → AI has a prompt for Codex, a button for Cursor, a button for the Claude app and JSON for the rest. This page is about Claude Code.
Will Claude Code publish my release notes?
No. It drafts them in Storewatch. You read the old and the new text and click Publish.
Source: Storewatch’s Settings → AI and MCP server, version 3.5 (November 2026). Claude Code’s own setup is in its MCP documentation. Checked November 2026.
Related: App Store Connect MCP server · ASO MCP server · AI agent · Claude and App Store Connect, setup guide · For indie developers
Made by Alex, the developer of Storewatch. Last updated November 21, 2026.