Create an AI Image with MCP
Call kp_capabilities first and select one complete entry from image.models.
Nano Banana 2
{ "type": "image", "prompt": "A clean binary-search course cover with a dark technical style", "image": { "provider": "gemini", "model": "nano-banana-2", "aspect_ratio": "16:9", "output_quality": "2K" }}
GPT Image 2
{ "type": "image", "prompt": "A binary-search poster for a developer course", "input_file_ids": ["file_xxx"], "image": { "provider": "azure_openai", "model": "gpt-image-2", "aspect_ratio": "3:2", "output_quality": "2K", "quality": "medium" }}
Notes
- provider and model must come from the same capabilities entry.
- aspect_ratio and output_quality must be supported by that bundle.
- quality is valid only for GPT Image.
- Upload reference images or documents with kp_upload_file.
- AI Image does not advertise URL Context; do not claim that it read a webpage.
- Poll with kp_get and call kp_outputs after ready.
