Provides access to Google's Gemini API for conversational AI interactions and image generation using Nano Banana, supporting multi-turn conversations, configurable context, and automated model selection for text-heavy images.
Gemini MCP Server
An MCP (Model Context Protocol) server that provides access to Google's Gemini API. Drop-in alternative to Codex MCP with matching interface.
Features
gemini - Start a new Gemini conversation with configurable context
gemini-reply - Continue multi-turn conversations
gemini-image - Generate images using Nano Banana (Gemini's native image generation)
Installation
Configuration
Create a .env file or set environment variables:
Get your API key from Google AI Studio.
Usage with Claude Code
Add to your MCP settings (~/.claude/mcp.json):
Tools
gemini
Start a new conversation with Gemini.
Parameters:
prompt(required) - The initial promptcwd- Working directory contextsandbox- Access policy: "read-only", "workspace-write", or "danger-full-access"base-instructions- Override default system instructionsdeveloper-instructions- Additional context for the model
Returns: Response text and a conversationId for follow-up.
gemini-reply
Continue an existing conversation.
Parameters:
conversationId(required) - ID from a previous gemini callprompt(required) - Your follow-up message
gemini-image
Generate images using Nano Banana, Google's native image generation built into Gemini.
Models:
Nano Banana (
gemini-2.5-flash-image) - Fast, cost-effective (~$0.04/image), good for most use casesNano Banana Pro (
gemini-3-pro-image-preview) - Advanced model with better text rendering, ideal for infographics, diagrams, and text-heavy images
Auto-Detection: The server automatically uses Nano Banana Pro when your prompt contains keywords like:
"nano banana pro", "pro model"
"infographic", "diagram", "chart", "graph"
"text", "typography", "font", "lettering"
"logo", "brand", "poster", "flyer", "banner"
"slide", "presentation", "document"
"high quality", "4k", "detailed text"
Parameters:
prompt(required) - Description of the image to generatenumberOfImages- How many images (1-4, default: 1)aspectRatio- Image ratio: "1:1", "3:4", "4:3", "9:16", "16:9"usePro- Force Nano Banana Pro (auto-detected from prompt if not specified)outputPath- Directory to save generated images
Example:
Development
License
MIT