# MCP Imagenate [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/mimo-3/mcp-imagenate  
**GitHub Stars:** 16  
**npm Downloads (last month):** 1382  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/mcp-imagenate

## Description
MCP server for image generation using multiple providers (Google Gemini, OpenAI, BFL FLUX)

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "mcp-imagenate": {
    "command": "npx",
    "args": ["-y","mcp-imagenate"],
    "env": {
      "GEMINI_API_KEY": "",
      "NANO_BANANA_API_KEY": "",
      "OPENAI_API_KEY": "",
      "GPT_IMAGE_API_KEY": "",
      "BFL_API_KEY": "",
      "REVE_API_KEY": "",
      "REVE_API_TOKEN": ""
    }
  }
}
```

**Requires environment variables:** `GEMINI_API_KEY`, `NANO_BANANA_API_KEY`, `OPENAI_API_KEY`, `GPT_IMAGE_API_KEY`, `BFL_API_KEY`, `REVE_API_KEY`, `REVE_API_TOKEN` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What MCP Imagenate does

MCP Imagenate exposes image generation through a single MCP interface while allowing the configured provider and model to vary by request. Supported image providers are Google Gemini, OpenAI gpt-image, BFL FLUX, and Reve. Google Gemini also provides a separate video-generation path through Gemini Omni, producing short clips with audio and on-screen text.

The available image model names include `nano-banana-2`, `nano-banana-2.1`, and `nano-banana-pro` for Gemini; `gpt-image-2.5-flare`, `gpt-image-2.5-sunburst`, and `gpt-image-2` for OpenAI; `flux-2-klein`, `flux-2-pro`, and `flux-2-max` for BFL; and `reve-image` for Reve. The default image model is `gpt-image-2.5-flare`.

## How it works

An MCP client starts the server as a Node.js process, and the server makes requests to the providers whose API keys are configured. Generated files are written to the selected output directory. The image tool can also read local input images for supported provider workflows, allowing prompts to include reference or editing inputs.

`generate_image` accepts a prompt, model, resolution, aspect ratio, generation mode, background setting, thinking mode, output directory, and optional input image paths. Supported aspect ratios range from square and portrait formats to `21:9`; resolutions are `1K`, `2K`, and `4K`. Google models can return an image with accompanying text when `mode` is `image_and_text`.

`generate_video` uses the Gemini API key and the `gemini-omni-1.1-flash` model for clips from 3 to 10 seconds. The image and video capabilities use the same Google credential when Gemini is selected.

## Setup and configuration

Node.js 22 or newer is required. Run the package directly with `npx mcp-imagenate`, or install it globally with npm. Set at least one provider credential before starting the process:

- Gemini: `GEMINI_API_KEY` or `NANO_BANANA_API_KEY`
- OpenAI: `OPENAI_API_KEY` or `GPT_IMAGE_API_KEY`
- BFL: `BFL_API_KEY`
- Reve: `REVE_API_KEY` or `REVE_API_TOKEN`

The alternative key names take precedence as documented for Gemini and OpenAI; `REVE_API_KEY` takes precedence over `REVE_API_TOKEN`. `NANO_BANANA_OUTPUT_DIR` optionally defines a base directory for saved images and sandboxes input and output paths within it. This is useful when restricting file access in production.

For Claude Desktop, add an `mcp-imagenate` entry that runs `npx` with `mcp-imagenate` as its argument and place the required credentials in the entry's `env` object.

## Tools and capabilities

- Generate images with Gemini, OpenAI, BFL FLUX, or Reve models.
- Generate 3–10 second Gemini Omni video clips with audio.
- Choose image resolution, aspect ratio, and output directory.
- Send local reference images to supported image providers.
- Request transparent PNG output with OpenAI gpt-image models.
- Use Reve for typography and layout-focused generations.

## Limitations and notes

Transparent backgrounds are supported only by OpenAI gpt-image models. Selecting transparency with Gemini, BFL FLUX, or Reve is rejected before the provider request is sent, and adding the phrase to the prompt does not change that limitation.

Reve ignores the requested resolution and returns its own large output. Its prompts are limited to 4,000 characters, and it accepts no more than eight reference images. Reve output may be PNG, JPEG, or WebP, with the saved extension matching the returned file format. A Reve generation costs 150 credits, typically takes 40–80 seconds, and should be given at least a 120-second timeout when called through a proxy or job runner.

_Full upstream README: https://allmcps.com/mcp/mcp-imagenate/readme_