Arenukvern/mcp_flutter

MCP Toolkit for Flutter AI Agent Driven Development (MCP/CLI + custom client side tools) - via closed feedback loop (visual & semantic snapshot) and high client side customization adaptable for any Flutter app. Nowadays it is often called as agentic harness.

What it solves

flutter-mcp-toolkit provides a way for AI agents (such as Claude Code, Cursor, and Zed) to interact with and control a running Flutter application. It eliminates the need for AI assistants to guess the state of the app or manually edit code without seeing the results, creating a closed feedback loop between the agent and the running software.

How it works

The project consists of a Dart MCP server and a Flutter package. By adding the toolkit to a Flutter app and running it in debug mode, an AI agent can connect to the app via the Model Context Protocol (MCP). The toolkit exposes a set of tools (prefixed with fmt_*) that allow the agent to perform actions like taking semantic snapshots, tapping widgets, typing into forms, and triggering hot-reloads.

Additionally, the toolkit supports Dynamic Tools Registration, allowing developers to define custom MCP tools and resources directly within the Flutter app that the AI agent can use at runtime.

Who it’s for

Flutter developers who want to use AI agents to automate debugging, testing, and UI development by giving the agent direct control over the running application.

Highlights

  • Direct App Control: Agents can tap, scroll, type, and navigate within the running app.
  • Deep Inspection: Ability to take semantic snapshots, read logs, and evaluate Dart expressions.
  • lifecycle Management: Support for hot-reload and hot-restart directly from the AI chat.
  • Dynamic Tooling: Apps can register their own custom MCP tools at runtime for the agent to use.
  • Automation Chain: A unified AutomationDriver contract that allows the same control vocabulary to work across MCP, CLI, and programmatic Dart scenarios.

Related

  • Project
    leancodepl/marionette_mcp

    An MCP server that enables AI agents to inspect and interact with running Flutter applications for automated smoke testing and interaction.

  • Project
    tidewave-ai/tidewave_phoenix

    An MCP server that gives AI coding agents runtime-level access to running Phoenix applications for code execution, log reading, and database querying.

  • Project
    Buoy-gg/buoy

    Buoy is a proprietary in‑app dev‑toolkit for React Native, Expo, Flutter, Capacitor/Ionic and Three.js. It adds a floating menu that lets developers inspect and modify network calls, storage, state caches, routes, render performance, and even 3‑D scenes while the app runs. The core package (`@buoy-gg/core`) is extensible with optional tools (network, storage, time‑machine, Ask Buoy, etc.). Tools can be synced to Buoy Desktop or an AI‑driven MCP server. Installation is via npm, followed by a quick‑start guide and a free or Pro Buoy account. The software is proprietary.

  • Project
    joshuayoes/ios-simulator-mcp

    An MCP server that enables AI agents to control iOS simulators, perform UI interactions, and automate QA testing through programmatic tool calls.