AssetGenie 🧞‍♂️

A complete, local Model Context Protocol (MCP) server written in pure Dart to automate, audit, and optimize Flutter project assets and localization structure. Connect it to AI clients like Cursor or Claude Desktop to let the AI interact directly with your project's assets and translation structures.

🚀 Features

AssetGenie exposes five powerful tools to the LLM context:

1. assetgenie_audit_health

Scans the project's pubspec.yaml and physical folders to identify asset anomalies:

• Flags missing physical files declared in pubspec.yaml.
• Flags physical assets that are not declared in pubspec.yaml (undeclared assets).
• Identifies heavy assets exceeding a configurable file size threshold (default: 500 KB).
• Automatically checks for unused physical files by searching for references in lib/ source code.

2. assetgenie_sync_localization

Safely updates or injects translation keys into local Application Resource Bundle (.arb) files:

• Appends new translations without manual copying/pasting.
• Standardizes formatting and groups metadata keys (@key) next to their respective properties.
• Alphabetizes keys according to standard ARB guidelines.

3. assetgenie_generate_constants

Auto-generates clean, modern Dart constants from physical asset files:

• Recursively walks through the physical assets folders.
• Converts file paths into structured camelCase variable names.
• Resolves identifier collisions and reserved keywords dynamically.
• Writes a safe config class Assets inside lib/generated/ to prevent typos and hardcoded string values.

4. assetgenie_audit_localization

Audits the project's ARB localization files to ensure quality and consistency:

• Compares target locales against a primary locale to find missing translation keys.
• Detects extra translation keys defined in target locales but missing in the primary locale.
• Validates that parameterized placeholders (e.g., {name}) match between translations to prevent runtime formatting crashes.
• Scans Dart code recursively in lib/ to identify unused translation keys.

5. assetgenie_app_harness

Connects to a running Flutter application via its Dart VM Service to inspect layout, capture state, or stream logs:

• Retrieves a clean, structured JSON Widget Summary Tree from the active application context.
• Requests a live PNG screenshot of the application's current layout, returning it directly in the tool response.
• Streams and captures stdout/stderr console prints and runtime logging statements over WebSocket.

🛠️ Installation & Setup

Ensure you have Dart SDK installed on your system.

1. Clone the repository:

   hljs language-bash

   Copy

   git clone https://github.com/aditzeb/AssetGenie.git
   cd AssetGenie
   

2. Download package dependencies:

   hljs language-bash

   Copy

   dart pub get
   

3. Run the verification/tests to make sure everything is ready:

   hljs language-bash

   Copy

   dart test test/assetgenie_test.dart
   

🔌 Connecting to LLMs & IDEs (MCP Clients)

To use AssetGenie inside your favorite AI assistant or development environment, configure it as a local stdio MCP server.

1. Antigravity IDE

Antigravity automatically discovers MCP servers configured in its global MCP configuration file:

• Windows: C:\Users\<Your-Username>\.gemini\antigravity-ide\mcp_config.json
• macOS / Linux: ~/.gemini/antigravity-ide/mcp_config.json

Add the following under the "mcpServers" object:

{
  "mcpServers": {
    "assetgenie": {
      "command": "dart",
      "args": [
        "run",
        "C:\\path\\to\\AssetGenie\\bin\\main.dart"
      ]
    }
  }
}

2. JetBrains IntelliJ IDEA / Android Studio

To bring AssetGenie's capabilities into IntelliJ IDEA or Android Studio, you can use the Continue plugin or the Cline / Roo Cline plugins.

Option A: Using Continue (via config.yaml)

Add the following to your global Continue configuration (~/.continue/config.yaml or C:\Users\<Your-Username>\.continue\config.yaml) under the mcpServers block:

mcpServers:
  - name: assetgenie
    type: stdio
    command: dart
    args:
      - "run"
      - "C:\\path\\to\\AssetGenie\\bin\\main.dart"

[!NOTE] Make sure to switch Continue to Agent Mode in your chat panel to enable tool execution.

Option B: Using Cline / Roo Cline

Add the following to your settings file (roo_cline_mcp_settings.json):

{
  "mcpServers": {
    "assetgenie": {
      "command": "dart",
      "args": [
        "run",
        "C:\\path\\to\\AssetGenie\\bin\\main.dart"
      ]
    }
  }
}

3. VS Code

You can use the Continue, Cline, or Roo Cline extensions in VS Code by applying the same configurations as JetBrains IDEs.

• Cline settings are stored in C:\Users\<Your-Username>\AppData\Roaming\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json (Windows) or ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json (macOS).

4. Claude Desktop

Add the following to your claude_desktop_config.json configuration file:

• Windows: C:\Users\<Your-Username>\AppData\Roaming\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "assetgenie": {
      "command": "dart",
      "args": [
        "run",
        "C:\\path\\to\\AssetGenie\\bin\\main.dart"
      ]
    }
  }
}

5. Cursor / Windsurf

Go to Settings -> Features -> MCP, click Add New MCP Server:

• Name: AssetGenie
• Type: stdio
• Command: dart run C:\path\to\AssetGenie\bin\main.dart

📖 Tool Schema Details

1. assetgenie_audit_health

• project_path (string, required): Absolute path to the Flutter root directory.
• max_size_kb (number, optional): Size threshold in KB (default: 500).

2. assetgenie_sync_localization

• project_path (string, required): Absolute path to the Flutter root directory.
• locale (string, required): Target language code (e.g. 'en', 'de').
• kv_pairs (object, required): A JSON map of translation pairs (e.g., {"loginButton": "Sign In"}).

3. assetgenie_generate_constants

• project_path (string, required): Absolute path to the Flutter root directory.
• output_filename (string, optional): Target file name (default: 'generated_assets.dart').

4. assetgenie_audit_localization

• project_path (string, required): Absolute path to the Flutter root directory.
• primary_locale (string, optional): The base locale to compare other locales against (default: 'en').
• unused_keys_check (boolean, optional): Whether to check for unused keys in lib/**/*.dart (default: true).

5. assetgenie_app_harness

• vm_service_uri (string, required): The Dart VM Service WebSocket or HTTP URI (e.g., http://127.0.0.1:8181/T9n3k_04gA=/).
• action (string, required): The instrumentation action to perform on the running Flutter application ('get_widget_tree', 'capture_screenshot', 'get_logs').

📄 License

This project is open-source and licensed under the MIT License. See LICENSE for more details.