# Use AppDoctor with AI

> Let Claude Code, Codex, Gemini CLI or a model in Ollama work on your keywords, listing, screenshots and videos, inside AppDoctor or from Claude, Cursor and other MCP apps.

AppDoctor works with the AI you already have. There's no second AI subscription and nothing new to sign up for: it uses the sign-in your AI tools already have on your computer, or a model running in Ollama.

There are two ways to use it.

## Inside AppDoctor

Click **AI** in the header of **Screenshots** or **Localization**. A chat opens beside the screen you're on.

1. Under the message box, choose the **project folder** with your app's code. The AI reads it to learn what your app really does, and runs in it, so your project's `CLAUDE.md` or `AGENTS.md` applies. AppDoctor suggests a folder when it finds one named after your app.
2. Choose the **AI and its model**. AppDoctor finds Claude Code, Codex (including the copy inside the ChatGPT app), Gemini CLI and Ollama with its models.
3. Choose the **effort**, for models that offer it: more effort is slower but more careful.
4. Type what you want, or pick a suggestion, and press Return.

The chat shows what the AI is doing in plain words ("Read your listing", "Drew a picture"), and you see each change on the canvas as it's made. ⌘Z undoes it. Follow-up messages carry on the same conversation.

## From Claude, Cursor and other apps

AppDoctor includes an MCP server, so AI apps on your computer can use its tools. In **Screenshots**, click **AI agent**: AppDoctor lists the AI apps it finds (Claude, Claude Code, Cursor, VS Code, Windsurf, Codex, Gemini CLI, Antigravity and LM Studio) and **Connect all** adds AppDoctor to each one's settings. It keeps everything else in those files and saves a copy first.

To add it by hand, the server is the AppDoctor app itself, run with `mcp`:

```json
{
  "mcpServers": {
    "appdoctor": {
      "command": "/Applications/AppDoctor.app/Contents/MacOS/appdoctor",
      "args": ["mcp"]
    }
  }
}
```

It speaks MCP over stdio and opens no network port. It only reads and writes AppDoctor's own files on your computer.

## What the AI can do

- **Keywords and listing:** read the keywords you track with their ranks, popularity and difficulty in each country, read your App Store listing in every language, run an [ASO audit](/docs/aso-audit), and suggest new listing text.
- **Screenshots:** add slides and layers, set backgrounds and themes, place any iPhone or iPad, and draw illustrations as SVG that come in as separate layers you can move and edit.
- **Videos:** turn a design into an app preview, add camera moves, callouts and sound, look at a storyboard and check it before exporting.
- **The iOS Simulator:** take screenshots and screen recordings of your app to put in device frames.

## What it can't do without you

- Listing suggestions never reach the App Store by themselves. They wait in **Localization**, marked with a sparkle. **Review** puts the text in the fields to edit, and **Review and Save…** sends it with your App Store Connect key after a confirmation that lists every change.
- Submitting for App Review is done in **App Review**, after a summary of the app, version and build that you confirm.

## Things to ask

- "Improve my listing for the keywords I track."
- "Which tracked keywords is my listing missing, and what would you change?"
- "Write listings for the other countries I track."
- "Make a 20-second app preview of the app's best features."
- "Draw an illustration of what the app does on slide 1."