Ask UI lets you point at a real Flutter screen and send that exact UI context to a coding agent.
Instead of describing a button, card, or layout problem in words, you open your app in Ask UI, select the widget on screen, add a comment, and let the agent work from the selected Flutter UI target.
ask_ui_example.webm
- Opens your running Flutter app in a browser-based workbench.
- Lets you select the UI element you want the agent to change.
- Captures comments and selected widget context together.
- Sends that context to Codex, Claude Code, pi, or another coding agent through Ask UI Chat.
- Uses
ask_ui_runtimeandask_ui_bridgeunder the hood to connect the workbench to your Flutter app.
Flutter app targets supported by Ask UI:
| Target platform | Android | iOS | Desktop |
|---|---|---|---|
| Device | Supported | Coming soon | Not supported |
| Emulator / simulator | Supported | Coming soon | Not supported |
Install the Ask UI bridge CLI:
dart pub global activate ask_ui_bridgeIn the Flutter project you want to inspect, add the runtime package:
flutter pub add ask_ui_runtimeRegister the runtime before runApp:
import 'package:ask_ui_runtime/ask_ui_runtime.dart';
void main() {
registerAskUiRuntime();
runApp(const MyApp());
}Install the Ask UI agent skill in your coding-agent environment:
npx skills add https://github.com/drown0315/ask_ui/tree/main/skillsThen ask your agent to start Ask UI from the Flutter project. If your app needs
a specific device, flavor, target, or Dart define, tell the agent in that same
request. The skill uses the published ask_ui_bridge CLI, starts the Flutter
app, handles device selection when needed, opens the workbench, and keeps
polling for Ask UI Chat messages.
- Dart SDK compatible with
ask_ui_bridge. - Flutter project running in debug mode.
ask_ui_runtimeregistered beforerunApp.- A supported target from the platform table above. Android targets must be available to Flutter and ADB.
The published user workflow is in skills/ask-ui. It tells the agent how to:
- start Ask UI from a Flutter project,
- handle bridge launch output and device selection,
- run the returned Agent Session Command,
- process Ask UI Chat messages with selected widget context,
- reply back into the Ask UI workbench.
The repository also contains skills/ask-ui-dev for Ask UI maintainers working
from this source checkout.