AssetGen MCP Server

AI-powered image asset generation with prompt engineering guidance and vectorization

Documentation

Quick Start (npx)

Run the MCP server directly without global installation:

npx -y @morphkit/asset-gen-mcp@latest

For MCP client configuration, use:

{
  "mcpServers": {
    "asset-gen": {
      "command": "npx",
      "args": ["-y", "@morphkit/asset-gen-mcp@latest"],
      "env": {
        "GOOGLE_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

For Claude Code:

claude mcp add asset-gen -e GOOGLE_API_KEY=YOUR_API_KEY -- npx -y @morphkit/asset-gen-mcp@latest

Features

• get-image-generation-prompt-instructions: Returns asset-type-specific prompt engineering best practices and examples for Google Imagen
• generate-images: Generates images using Google Imagen 4.0 with auto-vectorization for SVG outputs
• vectorize-image: Converts raster images to SVG using configurable vectorization options
• remove-background: AI-powered background removal using transparent-background

Requirements

• Node.js >= 18
• Python (for background removal) - Pre-installed on macOS and most Linux distributions
• Google API Key - Get one at Google AI Studio

First Run Notice

On first use, the remove-background tool downloads an ML model (~150MB) and creates a Python virtual environment. This takes approximately 3 minutes. Subsequent runs are fast (~2-5 seconds).

Installation

```bash cd packages/asset-gen-mcp bun install ```

Configuration

Copy the example environment file and configure:

```bash cp .env.example .env

Edit .env with your values

```

Development

```bash bun run dev ```

Production

Build

```bash bun run build ```

Run

```bash bun run start ```

Tools

get-image-generation-prompt-instructions

Returns asset-type-specific prompt engineering best practices and examples for Google Imagen

Usage:

```json { "name": "get-image-generation-prompt-instructions", "arguments": { "param": "value" } } ```

generate-images

Generates images using Google Imagen 4.0 with auto-vectorization for SVG outputs

Usage:

```json { "name": "generate-images", "arguments": { "param": "value" } } ```

vectorize-image

Converts raster images to SVG using configurable vectorization options

Usage:

```json { "name": "vectorize-image", "arguments": { "param": "value" } } ```

remove-background

AI-powered background removal using InSPyReNet model. Works with complex backgrounds, shadows, and lighting effects. Outputs transparent PNG.

Usage:

```json { "name": "remove-background", "arguments": { "inputPath": "/path/to/image.png", "outputPath": "/path/to/output.png", "fast": false } } ```

Set fast: true for quicker processing with slightly lower quality (384x384 model vs 1024x1024).

Testing

```bash bun run test ```

Type Checking

```bash bun run check-types ```

Linting

```bash bun run lint ```