Imagen

You run a multi-provider MCP server that intelligently selects the best image generation provider for each prompt, supports multiple backends, and saves results locally. You can tailor usage with per-provider models, reference images, and optional real-time data to produce consistent, high-quality visuals across sessions.

How to use

Install and run the server locally, then connect your MCP client to start generating images. The server automatically analyzes prompts and chooses the most suitable provider (OpenAI GPT-Image-1 or Gemini Nano Banana Pro). You can override the choice with a provider parameter when you need deterministic behavior.

How to install

Prerequisites: Python 3.11+ and a working network connection for API access.

Clone the project, install dependencies, and prepare the wrapper script.

git clone https://github.com/michaeljabbour/imagen-mcp.git
cd imagen-mcp
pip install -r requirements.txt
chmod +x run.sh

Configuration and usage notes

The server can be run locally via the provided wrapper script. Use the stdio MCP configuration to connect with a client.

{
  "mcpServers": {
    "imagen": {
      "type": "stdio",
      "command": "/path/to/imagen-mcp/run.sh",
      "args": [],
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "GEMINI_API_KEY": "AI..."
      }
    }
  }
}

Additional features and behavior

Auto Provider Selection analyzes prompts to pick the right backend, saving generated images to a default location or a custom path you specify. Gemini supports reference images (up to 14) and Google Search grounding for real-time data, while OpenAI excels with text-heavy visuals and diagrams.

Default save location is a local directory such as ~/Downloads/images/ with per-provider subfolders. You can customize the output path using output_path and override the base directory with the OUTPUT_DIR environment variable.

Usage examples

Auto provider selection example: generate an infographic with a prompt that benefits from combined text and visuals.

Manual provider override example: generate_image(prompt="...", provider="openai") or generate_image(prompt="...", provider="gemini").

Available tools

generate_image

Main tool with auto provider selection for image generation

conversational_image

Multi-turn refinement with history to iteratively improve images

list_conversations

List active conversations and their history

list_providers

Show available providers and capabilities

list_gemini_models

Query available Gemini image models