flutter-marionette

flutter-marionette

LeanCode Flutter plugin for Marionette MCP — runtime interaction with a live Flutter app via an MCP server, designed for AI-driven exploration, smoke verification, and UI debugging.

Included assets

• skills/flutter-marionette-usage/references/marionette.md — Marionette vs Patrol, setup, binding, MCP tools, and runtime workflow
• skills/flutter-marionette-usage/references/marionette-widget-config.md — MarionetteConfiguration, custom widgets, and screenshot sizing
• skills/flutter-marionette-usage/SKILL.md — explains what this plugin does, when to use it, and which rule or skill to reach for next
• .mcp.json — shipped marionette MCP launcher for plugin installs in Claude Code

Marionette vs Patrol

	Marionette (this plugin)	Patrol (flutter-patrol)
Purpose	Runtime exploration, smoke verification	Deterministic E2E test suites
Runs against	Live flutter run debug session	patrol develop / patrol test
Test files	None — agent drives the app	Dart test files in patrol_test/
Best for	Iterating on a feature, smoke after refactor	Regression-proof suites in CI
Build mode	Debug only	Debug + release

Both plugins can coexist; they solve different parts of the AI-assisted testing workflow.

Setup highlights

1. Install the MCP server

For the bundled plugin launcher, the lowest-friction setup is a global install:

dart pub global activate marionette_mcp

Alternative: add it as a dev dependency and run it with dart run marionette_mcp from your own client-specific MCP launcher.

2. Add the Flutter package to your app

flutter pub add marionette_flutter

3. Initialize the binding in main.dart

import 'package:flutter/foundation.dart';
import 'package:flutter/widgets.dart';
import 'package:marionette_flutter/marionette_flutter.dart';

void main() {
  if (kDebugMode) {
    MarionetteBinding.ensureInitialized();
  } else {
    WidgetsFlutterBinding.ensureInitialized();
  }
  runApp(const MyApp());
}

For apps with a custom design system, pass a MarionetteConfiguration — see skills/flutter-marionette-usage/references/marionette-widget-config.md.

Important: MarionetteBinding must be the only binding initialized in the process. If tests call main() in debug mode, avoid initializing Marionette in tests by checking FLUTTER_TEST or by using a separate test entrypoint.

4. Use the bundled MCP launcher

When this plugin is installed in Claude Code, it ships the marionette MCP server entry for you. You should not need to hand-edit MCP config just to enable Marionette.

The bundled launcher runs:

{
  "mcpServers": {
    "marionette": {
      "command": "marionette_mcp",
      "args": []
    }
  }
}

That means the remaining manual setup is:

• install marionette_mcp
• add marionette_flutter to the app
• initialize MarionetteBinding

5. Manual fallback for other MCP clients

If your MCP client does not load plugin-bundled servers, copy the same launcher into that client's MCP config:

{
  "mcpServers": {
    "marionette": {
      "command": "marionette_mcp",
      "args": []
    }
  }
}

6. Run and connect

flutter run

Copy the VM service URI from the run output (format: ws://127.0.0.1:PORT/ws). In your AI agent, call the connect tool with that URI.

Log collection

get_logs requires a configured LogCollector.

• Use LoggingLogCollector() for apps using the logging package.
• Use LoggerLogCollector() for apps using the logger package.
• Use PrintLogCollector() for custom logging setups.

Build-mode constraint

Marionette relies on Flutter's VM Service and is intended for a live flutter run session. It does not work in release builds. Use Patrol for release-mode automation.

Example usage

• /flutter-marionette-usage — get a short explanation of what this plugin does, how to use it, and example next requests

Related plugins

• flutter-patrol — deterministic E2E testing with Patrol MCP
• flutter-ui — design-system guidance (relevant when configuring isInteractiveWidget for custom widgets)