Marionette MCP Server

Flutter application UI interaction for AI agents via Model Context Protocol.

Data last scanned today · Reviewed today

45 1 dangerous tool13 tools★ 364DartApache-2.0

Overview

Marionette MCP acts as a bridge between AI agents and running Flutter applications, enabling automated interaction with UI elements. It allows models to navigate the widget tree, input text, trigger taps, perform scrolls, and capture screenshots to facilitate AI-driven smoke testing.

Our verdict

This server carries a high-risk rating with a security score of 45/100. Key concerns include that OAuth support and read-only mode have not yet been editorially reviewed, and the project is maintained by the community rather than the official vendor. Developers should exercise caution when integrating this tool into their workflows until security audits are completed.

Zero direct dependencies minimize supply-chain surface area.
Specifically architected for runtime Flutter UI interactions.
Contains 21 active contributors.
Includes documentation for setup and integration.

Authentication methods have not undergone security review.
Lacks an official homepage link for verification.
High-risk score due to missing audits.

Tools

Tool	Description	Risk
get_interactive_elements	Inspect the widget tree to find interactive elements.	low
tap	Perform a tap action on an element.	medium
secondary_tap	Perform a secondary tap action on an element.	medium
double_tap	Perform a double tap action on an element.	medium
long_press	Perform a long press action on an element.	medium
swipe	Perform a swipe action.	medium
pinch_zoom	Perform a pinch zoom action.	medium
scroll_to	Scroll to a specific target.	medium
enter_text	Enter text into an input field.	medium
press_back_button	Press the system back button.	medium

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What are the core requirements to use Marionette MCP with my app?

You must add the marionette_flutter package to your app and initialize MarionetteBinding in your main.dart file. Additionally, your Flutter application must be running in debug mode to provide the required VM service URI.

›Does Marionette MCP support apps built with custom design systems?

Yes, but custom design systems require additional configuration. Without specific setup, the AI agent may be unable to identify or interact with your custom buttons and fields.

›How does Marionette MCP differ from the official Flutter MCP server?

Marionette MCP is specifically designed for runtime interaction, such as tapping, scrolling, and automated smoke testing. The official Flutter MCP server focuses on development-time tasks like dependency management and code analysis.

›Can I perform actions in my app that aren't native UI interactions?

Yes, you can expose custom actions to the AI agent by implementing Custom Extensions. This allows you to perform app-specific tasks like seeding test data, navigating by route name, or toggling feature flags.