Skip to content
pubscout

marionette_mcp

0.7.0, released 7 weeks ago by leancode.co

MCP server allowing AI agents to inspect and interact with running Flutter applications.

$ flutter pub add marionette_mcp
Source repository

What it does

Bridges AI agents like Claude, Copilot, and Cursor directly to running Flutter apps via the Model Context Protocol (MCP). Enables agents to view widget trees, tap elements, input text, scroll, and capture screenshots for automated testing and interaction.

AI-written summary of version 0.7.0, based on the package's README and metadata. Report an inaccuracy

Use it when

  • You want AI agents to interact with and smoke test running Flutter apps.
  • You need to automate UI inspection, tapping, text entry, and screenshots via LLMs.

Skip it when

  • You are looking for a development-time tool to search pub.dev or analyze static code.
  • You cannot make minor modifications to initialize MarionetteBinding in your app's main function.

Strengths

  • Connects AI agents directly to running applications through MCP.
  • Provides a small, focused toolset optimized for token-efficient agent context.
  • Supports custom extensions to expose app-specific actions as agent tools.

Weaknesses

  • Requires manual initialization and binding setup in the target Flutter application.
  • Custom design systems require explicit configuration so agents can recognize custom widgets.

Setup

Kind
Pure Dart
Effort
Medium
Code generation
No
Direct dependencies
6

Add marionette_flutter dependency, initialize MarionetteBinding in main.dart, activate marionette_mcp globally, and configure your AI tool.

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

Alternatives

Packages that do the same job and could replace marionette_mcp.

Works with

Packages commonly used together with marionette_mcp: ones it builds on, ones built on it, and its own family.