mcp-image and nanobanana-mcp
These two tools appear to be competing implementations of an MCP server for AI image generation using Gemini Vision and Nano Banana models, with "shinpr/mcp-image" offering additional prompt optimization and quality presets, while "YCSE/nanobanana-mcp" specifically targets Claude Desktop and Claude Code integration.
About mcp-image
shinpr/mcp-image
MCP server for AI image generation and editing with automatic prompt optimization and quality presets (fast/balanced/quality). Powered by Gemini (Nano Banana 2 & Pro).
Uses Gemini 2.5 Flash for automatic prompt optimization via a Subject–Context–Style framework, enriching simple descriptions with photographic and artistic details while preserving intent. Implements three quality tiers (Nano Banana 2 & Pro) for generation, integrates with MCP-compatible tools (Cursor, Claude Code, Codex), and includes an optional standalone Agent Skill for teaching AI assistants better prompt-writing practices without requiring API access.
About nanobanana-mcp
YCSE/nanobanana-mcp
Gemini Vision & Image Generation MCP for Claude Desktop and Claude Code
Provides image editing alongside generation, session-aware context preservation across multiple turns, and runtime model switching between Flash (faster) and Pro (higher-quality) variants without server restart. Implements MCP server architecture with stdio transport, integrating Google's Gemini 3.1 image models directly into Claude Desktop, Claude Code, and Cursor via configurable JSON manifests.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work