Image Generation MCP Server

An MCP (Model Context Protocol) server that generates images using the laozhang.ai Gemini 3 Pro API and saves them directly to your Obsidian vault.

Features

• generate_image — Generate a single image from a text prompt
• generate_multiple — Generate 2-4 images from the same prompt (different interpretations)
• generate_variations — Transform reference image(s) with a style/modification prompt
• list_generated_images — List recently generated images

Images are automatically saved with descriptive filenames based on the prompt (e.g., 2026-01-19_cute_robot_waving.jpg).

Requirements

• Python 3.11+
• uv package manager
• laozhang.ai API key

Installation

git clone https://github.com/JasonBates/image-gen-mcp.git
cd image-gen-mcp
uv sync

Configuration

1. Copy the example environment file:

   cp .env.example .env

2. Edit .env with your settings:

   LAOZHANG_API_KEY=your-api-key-here
   OBSIDIAN_VAULT_PATH=~/path/to/your/vault
   AI_GRAPHICS_FOLDER=Assets/AI graphics

Usage

Adding to Claude Code

Add to your Claude Code MCP configuration (~/.claude.json):

{
  "mcpServers": {
    "image-gen": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/image-gen-mcp", "image-gen-mcp"],
      "env": {
        "LAOZHANG_API_KEY": "your-api-key-here"
      }
    }
  }
}

Testing with MCP Inspector

npx @modelcontextprotocol/inspector uv run image-gen-mcp

Tools

generate_image

Generate a single image from a text prompt.

Parameter	Default	Description
prompt	(required)	Text description of the desired image
aspect_ratio	"1:1"	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21, 5:4
size	"2K"	1K (fast), 2K (default), 4K (highest quality)

generate_multiple

Generate multiple fresh interpretations of the same prompt.

Parameter	Default	Description
prompt	(required)	Text description
count	4	Number of images (2-4)
aspect_ratio	"1:1"	See above
size	"2K"	See above
diversity	0.0	Prompt variation level (0.0-1.0). 0.0 = identical prompts, 0.3 = subtle changes, 0.6 = style/mood variations, 1.0 = creative reinterpretations

Example with diversity:

# Low diversity - subtle variations (lighting, minor details)
generate_multiple("a robot waving", count=4, diversity=0.3)

# High diversity - creative reinterpretations
generate_multiple("a robot waving", count=4, diversity=0.9)
→ "a vintage 1950s tin robot waving on a kitchen counter"
→ "a massive robot waving goodbye to a departing spaceship"
→ "a tiny robot waving from inside a snow globe"
→ "a robot made of flowers waving in a garden"

generate_variations

Transform reference image(s) using a style/modification prompt.

Parameter	Default	Description
image_paths	(required)	List of paths to reference images (1-14)
prompt	(required)	How to transform (e.g., "in watercolor style")
count	4	Number of variations (1-4)
aspect_ratio	"1:1"	See above
size	"2K"	See above

Example: Take an existing robot image and generate pixel art and cyberpunk versions.

list_generated_images

List recently generated images in the output folder.

Parameter	Default	Description
limit	20	Maximum images to list

API Capabilities

The laozhang.ai Gemini 3 Pro API supports:

Feature	Options
Aspect Ratios	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21, 5:4
Sizes	1K (~1024px), 2K (~2048px), 4K (~4096px)
Reference images	Up to 14 for style/content guidance

Note: The API generates one image per request. For multiple images, this server makes sequential requests automatically.

Filename Generation

Filenames are auto-generated from your prompt:

1. Key terms extracted (skipping common words)
2. Sanitized (lowercase, underscores)
3. Date prefix added

Examples:

• Single: 2026-01-19_cute_robot_waving.jpg
• Multiple: 2026-01-19_cute_robot_waving_v1.jpg, _v2.jpg
• Variations: 2026-01-19_watercolor_style_var1.jpg, _var2.jpg

License

MIT