MCP 服务器

PanPan Image Generator MCP Server

AI-powered image generation and editing server for Claude Code, supporting multiple Gemini models, batch generation, shared context, and conversion to PDF/PPTX.

README

<h1 align="center">PanPan Image Generator MCP</h1>

<p align="center"> <strong>AI-powered image generation & editing MCP server for Claude Code</strong> </p>

<p align="center"> <a href="#features">Features</a> • <a href="#supported-models">Models</a> • <a href="#quick-start">Quick Start</a> • <a href="#tools">Tools</a> • <a href="#examples">Examples</a> </p>

Features

Multi-model support — Gemini Pro Image Preview (1K/2K/4K) and Gemini Flash Image Preview
Image editing — Edit existing images with text instructions, supports multi-image reference
Batch generation — High-concurrency worker pool for parallel image generation
Shared context — Generate series of images with consistent style (great for slides, tutorials, card sets)
Resolution control — 1K / 2K / 4K presets or custom WxH, plus aspect ratio selection
Format conversion — Convert images to PDF or PowerPoint (PPTX)
Smart file handling — Auto-fallback output directory, customizable per-request save paths

Supported Models

Model ID	Alias	Quality
`gemini-3-pro-image-preview`	`nanobananapro`	High quality, 1K
`gemini-3-pro-image-preview-2k`	—	High quality, 2K
`gemini-3-pro-image-preview-4k`	—	High quality, 4K
`gemini-3.1-flash-image-preview`	`nanobanana2`, `gemini-flash`	Fast

All models are accessed through any OpenAI-compatible API endpoint that supports Gemini image generation.

Demo

<table> <tr> <td align="center"> <img src="assets/demo-art.png" width="300"><br> <sub>Art illustration</sub> </td> <td align="center"> <img src="assets/demo-cyberpunk.png" width="300"><br> <sub>Cyberpunk</sub> </td> <td align="center"> <img src="assets/demo-chinese.png" width="300"><br> <sub>Chinese ink painting</sub> </td> </tr> </table>

Quick Start

Install

git clone https://github.com/panda920/panpan-image-generator-mcp.git
cd panpan-image-generator-mcp
npm install

Configure

Copy the example env file and fill in your API key:

cp .env.example .env

Edit .env:

GEMINI_API_KEY=your_api_key_here
GEMINI_API_BASE=https://openrouter.ai/api/v1

Use with Claude Code

Add to ~/.claude.json:

{
  "mcpServers": {
    "panpan-image-generator": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/panpan-image-generator-mcp/panpan-image-generator-mcp-server.js"],
      "env": {
        "GEMINI_API_KEY": "your_api_key",
        "GEMINI_API_BASE": "https://openrouter.ai/api/v1"
      }
    }
  }
}

Tools

Image Generation

Tool	Description
`generate_image_batch`	Generate one or more images with high concurrency. Supports all 4 models, custom resolution and aspect ratio.
`generate_image_with_shared_context`	Batch generate images with shared style context for visual consistency. Ideal for slides, tutorials, card sets.

Image Editing

Tool	Description
`edit_image_nano`	Edit images (single or batch). Supports all 4 models, multi-image reference, custom resolution and aspect ratio.

Format Conversion

Tool	Description
`images_to_pdf`	Convert multiple images to a PDF document (one image per page, full bleed).
`images_to_pptx`	Convert multiple images to a PowerPoint presentation.

Examples

In Claude Code, just ask naturally:

Generate a cyberpunk cityscape at night in 4K

Edit this image: change the background to blue

Batch generate 5 cat illustrations in different styles

Create a 10-slide presentation with consistent branding, then export to PPTX

Environment Variables

Variable	Default	Description
`GEMINI_API_KEY`	—	API key (required)
`GEMINI_API_BASE`	`https://openrouter.ai/api/v1`	OpenAI-compatible API base URL
`GEMINI_MODEL`	`gemini-3-pro-image-preview`	Default Pro model
`GEMINI_FLASH_MODEL`	`gemini-3.1-flash-image-preview`	Default Flash model
`OUTPUT_DIR`	Current working directory	Default output directory