Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@NanoBanana MCP Generate an image of a cyberpunk city at sunset"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server that brings Gemini's image generation and editing capabilities to Claude Desktop, Claude Code, and Cursor. Supports Nano Banana 2 (Flash) and Nano Banana Pro models.
Image Generation - Create 2K images from text prompts
Image Editing - Transform images with natural language instructions
Session Consistency - Maintain style/character across generations
Runtime Model Switching - Switch between Flash and Pro models without restart
Multi-turn Chat - Conversational context with image support
Related MCP server: nb-mcp
Node.js 18+
Google AI API Key (Get one here)
claude mcp add nanobanana-mcp -- npx -y @ycse/nanobanana-mcp \
-e "GOOGLE_AI_API_KEY=your_api_key"Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"nanobanana-mcp": {
"command": "npx",
"args": ["-y", "@ycse/nanobanana-mcp"],
"env": {
"GOOGLE_AI_API_KEY": "your_api_key"
}
}
}
}Create or edit .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global):
{
"mcpServers": {
"nanobanana-mcp": {
"command": "npx",
"args": ["-y", "@ycse/nanobanana-mcp"],
"env": {
"GOOGLE_AI_API_KEY": "your_api_key"
}
}
}
}See Cursor MCP Documentation for more details.
Tool | Purpose |
| Required. Set aspect ratio before image generation |
| Switch between flash/pro models at runtime |
| Generate images from text prompts |
| Edit images with natural language |
| Multi-turn conversation with images |
| View session image history |
| Reset session context |
Must be called before generating or editing images.
Valid ratios: 1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 5:4, 4:5, 21:9Switch models per-session without restarting:
Value | Model | Description |
| gemini-3.1-flash-image-preview | Nano Banana 2 - Faster (default) |
| gemini-3-pro-image-preview | Nano Banana Pro - Higher quality |
{
prompt: string; // Image description
aspect_ratio?: string; // Override session ratio
output_path?: string; // Save path (default: ~/Documents/nanobanana_generated/)
conversation_id?: string; // Session ID
use_image_history?: boolean; // Use previous images for consistency
reference_images?: string[]; // Reference images for style
}{
image_path: string; // File path, "last", or "history:N"
edit_prompt: string; // Edit instructions
aspect_ratio?: string; // Override session ratio
output_path?: string; // Save path
conversation_id?: string; // Session ID
reference_images?: string[]; // Style references
}npx @ycse/nanobanana-mcp --install-commands claude-code
# Or manually:
# mkdir -p ~/.claude/commands
# cp commands/claude-code/*.md ~/.claude/commands/npx @ycse/nanobanana-mcp --install-commands cursor
# Or manually:
# mkdir -p .cursor/commands
# cp commands/cursor/*.md .cursor/commands/See Cursor Slash Commands for more details.
/nb-flash - Switch to Flash model (faster)
/nb-pro - Switch to Pro model (higher quality)1. Set aspect ratio: set_aspect_ratio("16:9")
2. Generate: "A cyberpunk cityscape at sunset"// First image
{ prompt: "A red-hat cat", conversation_id: "cat" }
// Second image - same character
{ prompt: "The cat taking a nap", conversation_id: "cat", use_image_history: true }// Edit the last generated image
{ image_path: "last", edit_prompt: "Change hat to blue" }
// Edit specific image from history
{ image_path: "history:0", edit_prompt: "Add sunglasses" }// Start with Flash for quick iterations
set_model({ model: "flash" })
{ prompt: "Draft concept art" }
// Switch to Pro for final render
set_model({ model: "pro" })
{ prompt: "Final polished version", use_image_history: true }Variable | Required | Description |
| Yes | Google AI API key |
| No | Default model ( |
Generated images save to ~/Documents/nanobanana_generated/:
Generated: generated_[timestamp].png
Edited: [original]_edited_[timestamp].png
git clone https://github.com/YCSE/nanobanana-mcp.git
cd nanobanana-mcp
npm install
npm run dev # Development mode with hot reload
npm run build # Production build
npm run start # Run compiled serverImage generation fails:
Verify API key is valid
Check quota at Google AI Studio
Ensure set_aspect_ratio was called first
Tools not showing:
Restart Claude Desktop/Code
Check config file syntax
Verify npx -y @ycse/nanobanana-mcp runs without errors
MIT
Clear conversation history for a specific conversation ID
Inputs: conversation_id (required).
Chat with Gemini 3.1 Flash model. Supports multi-turn conversations with up to 10 reference images.
Inputs: message (required), images, system_prompt, conversation_id.
Edit or modify existing images based on prompts. Supports session history references ('last' or 'history:N') and image consistency features.
Inputs: image_path (required), edit_prompt (required), output_path, aspect_ratio, +3 more.
Generate images using Gemini's image generation capabilities. Supports session-based image consistency for maintaining style/character across multiple generations.
Inputs: prompt (required), output_path, aspect_ratio, conversation_id, +3 more.
Get the list of generated/edited images in a session for reference
Inputs: conversation_id (required).
Set the aspect ratio for subsequent image generation and editing in this session. Must be called before generating/editing images if a specific ratio is desired.
Inputs: aspect_ratio (required), conversation_id.
Set the Gemini model for this session. 'flash' for faster generation (default), 'pro' for higher quality.
Inputs: model (required), conversation_id.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 7 tools
Each tool has a clearly distinct purpose: retrieving history, chatting, generating images, editing images, clearing conversation, and setting session parameters. There is no functional overlap between generate and edit, and chat is text-focused while image tools are visual. Even the history and clear tools are unambiguous (images vs. conversation).
Most tools follow a verb_noun pattern (get_image_history, clear_conversation, set_aspect_ratio, set_model), but three tools are prefixed with 'gemini_' (gemini_chat, gemini_generate_image, gemini_edit_image), creating an inconsistency in form. The verbs are clear, but the prefix usage is not uniform, and 'gemini_chat' reads more as a noun phrase than a verb_noun command.
Seven tools is a well-scoped count for a server focused on image generation/editing and session management. Each tool serves a necessary function without redundancy or bloat, fitting comfortably within the ideal 3-15 range.
The surface covers the main lifecycle for image work: generate (create), edit (update), view history (read), and session management. The only minor gap is the lack of an explicit delete/image removal tool and no way to view conversation history outside of image history, but these are workarounds in most use cases.
MCP server for Qwen Image 3 AI image generation
Shared memory MCP server for Claude Code, Codex, Gemini CLI and Cursor. Cited, editable.
Generate images from Claude, ChatGPT, Cursor and Codex with 12 top AI image models. Up to 4 images per call, SVG vectorization, background removal, upscaling and image-to-video. Remote MCP with OAuth, no API keys.
Focused MCP server for OpenAI image/audio generation (v2.0.0). Wraps endpoints via HAPI CLI.
