Generate Image MCP

Let AI generate and edit images directly in Claude Desktop, OpenCode, Cursor, and other MCP-compatible clients.

Features

• Multi-turn Dialogue - Collect user requirements step by step
• Smart Inference - Automatically infer missing parameters from context
• Session Management - Maintain conversation state across multiple calls
• No Auto-fill - Never fill in parameters without user confirmation

Quick Start

1. Install

npm install -g generate-image-mcp

2. Get an API key

Register at agnes-ai.com to get your API key (free quota available).

3. Configure

Add to your MCP config file:

OpenCode (~/.config/opencode/opencode.json):

{
  "mcp": {
    "generate-image": {
      "type": "local",
      "command": ["npx", "generate-image-mcp"],
      "enabled": true,
      "environment": {
        "AGNES_API_KEY": "your-api-key-here",
        "AGNES_API_URL": "https://apihub.agnes-ai.com/v1/images/generations"
      }
    }
  }
}

Claude Desktop (%APPDATA%\Claude\claude_desktop_config.json):

{
  "mcpServers": {
    "generate-image": {
      "command": "npx",
      "args": ["-y", "generate-image-mcp"],
      "env": {
        "AGNES_API_KEY": "your-api-key-here"
      }
    }
  }
}

Other clients:

| Client | Config file | |--------|-------------| | Claude Desktop | %APPDATA%\Claude\claude_desktop_config.json | | OpenCode | ~/.config/opencode/opencode.json | | Cursor | ~/.cursor/mcp.json | | Windsurf | ~/.windsurf/mcp.json | | Cline (VS Code) | Cline MCP settings UI |

4. Restart your client

Done! Ask your AI: "Generate an image of a sunset over mountains".

How It Works

Multi-turn Dialogue

When you ask to generate an image, the tool will:

1. Analyze your request - Extract what it can understand
2. Ask for missing info - Generate 4 options for you to choose
3. Collect your choices - Remember across multiple calls
4. Generate image - Only when all info is complete

Example:

User: Generate a garlic zombie image
AI: What type of scene? A) Character *recommended B) Abstract C) Animal D) Product
User: A
AI: What style? A) Realistic *recommended B) Anime C) Oil painting D) 3D
User: Realistic
AI: What background? A) Natural *recommended B) Indoor C) Urban D) Auto
User: A
→ Image generated!

Smart Inference

The tool can infer missing parameters from context:

• User selects "Anime style" → style = "anime"
• User selects "City night" → scene_type = "city" AND background = "urban"
• User selects "Neon lights" → lighting = "special effects"

Tools

generate_image

Generate an image from a text description.

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | raw_input | string | Yes | User's original input, passed directly without modification | | session_id | string | No | Session ID for multi-turn dialogue | | size | string | No | Image size (default: 1024x1024) | | model | string | No | Model name (default: agnes-image-2.1-flash) | | n | number | No | Number of images (default: 1) |

repaint_image

Edit an existing image by describing what to change.

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | image_url | string | Yes | URL of the original image | | edit_prompt | string | Yes | Description of changes | | size | string | No | Image size (default: 1024x768) | | model | string | No | Model name (default: agnes-image-2.1-flash) | | strength | number | No | Edit strength 0-1 (default: 0.7) |

Configuration

Environment Variables

| Variable | Required | Default | Description | |----------|----------|---------|-------------| | AGNES_API_KEY | Yes | — | Your Agnes AI API key | | AGNES_API_URL | No | https://apihub.agnes-ai.com/v1/images/generations | API endpoint |

Alternative Providers

OpenAI DALL-E:

"environment": {
  "IMAGE_PROVIDER": "openai",
  "IMAGE_API_KEY": "sk-your-openai-key",
  "IMAGE_MODEL": "dall-e-3"
}

Stability AI:

"environment": {
  "IMAGE_PROVIDER": "stability",
  "IMAGE_API_KEY": "sk-your-stability-key"
}

API Key Sources

| Provider | Get a key at | |----------|-------------| | Agnes AI | agnes-ai.com | | OpenAI | platform.openai.com | | Stability AI | platform.stability.ai |

From Source (Developers)

git clone https://github.com/m1908293028-boop/generate-image-mcp.git
cd generate-image-mcp
npm install
npm run build

Then configure with the local path:

{
  "mcp": {
    "generate-image": {
      "type": "local",
      "command": ["node", "path/to/generate-image-mcp/dist/server.js"],
      "enabled": true,
      "environment": {
        "AGNES_API_KEY": "your-api-key-here"
      }
    }
  }
}

Changelog

v1.2.0

• Added multi-turn dialogue with session management
• Added smart inference for missing parameters
• Removed auto-fill behavior
• Added option letter parsing (A/B/C/D)
• Added deduplication logic in prompt building

v1.1.x

• Initial release with basic image generation

License

MIT