---

name: nanobanana description: AI image generation using Google Gemini models. Generate, edit, and refine images from text prompts. license: MIT

Nanobanana Image Generation

Generate and edit images using Google Gemini's image generation capabilities through the Nanobanana MCP server.

Overview

This skill provides access to Gemini's native image generation and editing features via a local MCP server. Images are generated based on text prompts and can be iteratively refined.

Capabilities

generate_image

Create new images from text descriptions.

Generate a photorealistic image of a mountain landscape at golden hour

edit_image

Modify existing images with natural language instructions.

Edit this image: replace the cloudy sky with a starry night

continue_editing

Apply additional changes to the most recently generated/edited image.

Continue editing: add northern lights in the sky

MCP Configuration

The skill bundles the Nanobanana MCP server. Tools are automatically available when this skill loads.

Server path: Set NANOBANANA_MCP_PATH to your nanobanana-mcp-cliproxyapi installation.

Usage Examples

Basic Generation

Create an image of a cozy coffee shop interior with warm lighting

Styled Generation

Generate a watercolor painting of cherry blossoms over a Japanese garden

Photo Editing

Edit the uploaded image: remove the person in the background

Iterative Refinement

Generate a cyberpunk cityscape
Continue editing: add flying cars
Continue editing: make it rainy with neon reflections

Prerequisites

Environment Variables

Variable	Required	Description
NANOBANANA_MCP_PATH	Yes	Path to nanobanana-mcp-cliproxyapi
CLIPROXY_BASE_URL	Yes	Your cliproxy instance URL
CLIPROXY_API_KEY	Yes	API key for cliproxy
NANOBANANA_MODEL	Yes	Model: gemini-2.0-flash-exp or gemini-2.0-pro-exp

Setup

1. Clone and install nanobanana-mcp-cliproxyapi
2. Configure cliproxy with your API credentials
3. Set environment variables
4. Symlink this skill to ~/.config/agents/skills/nanobanana

See docs/SETUP.md for detailed instructions.

Output

Generated images are saved locally. The tool returns the file path to the generated image which can be viewed or further edited.

Tips

• Be specific about style, lighting, and composition
• Use iterative refinement for complex scenes
• Flash model for quick drafts, Pro model for final renders
• See docs/PROMPTS.md for prompt engineering guide