@wa008/ui-audit-mcp
v2.4.3
Published
MCP server for iOS app UI evaluation and testing, powered by idb + xcrun simctl
Maintainers
Readme
UI Audit MCP
Let AI agents evaluate your iOS app's UI — screenshots, taps, swipes, and a 3-dimension scoring system, all from your MCP client.
Features
- Device Control: Launch apps, take screenshots, tap, and swipe on iOS Simulators.
- UI Evaluation: 3-dimension user-perspective evaluation (outcome, usability, aesthetics).
Prerequisites
- macOS (required for iOS Simulator)
- Node.js 18+
- Xcode CLI Tools:
xcode-select --install - idb (required for tap/swipe):
brew install idb-companion pip3 install fb-idb
Install
Add to your MCP client config (Claude Desktop, Cursor, etc.):
{
"mcpServers": {
"ui-audit": {
"command": "npx",
"args": ["-y", "@wa008/ui-audit-mcp"]
}
}
}Tools
| Tool | Purpose |
|---|---|
| launch_app | Launch app by Bundle ID |
| take_screenshot | Capture screen & track evaluation step |
| tap / swipe | Interact with UI (ratio coordinates), screenshot, and track step |
| input_text | Type text into the focused field, screenshot, and track step |
| evaluate | Get dimension prompt or submit score + get next |
| get_audit_status | View missing evaluations or final markdown report |
Typical Workflow
launch_app("com.example.app")take_screenshot("MyTestCase", 1, "Verify Initial Screen")evaluate("MyTestCase", 1)→ Returns first dimension prompt- Agent analyzes UI → submits score → repeats until all dimensions done
get_audit_status(["MyTestCase"])→ Full markdown report
Evaluation Dimensions
Each step is scored on up to 3 dimensions (score ≥ 8 to pass):
- Outcome — Did the action produce the expected result? (action steps only)
- Usability — Is everything visible, understandable, and interactive?
- Aesthetics — Does it look professional and polished?
Data
- Logs:
~/.ui-audit-mcp/logs/ - Screenshots:
~/.ui-audit-mcp/screenshots/
License
MIT
