review_image
Check a generated image against its brief with a vision model to identify mismatches, garbled text, composition issues, and artifacts. Get a revised prompt to decide whether to regenerate the image.
Instructions
Have a vision model look at an image and critique it against a brief. Returns what matches, what is wrong (garbled text, composition, colors, artifacts) and a revised prompt. Use it after generate_image to check the result and decide whether to regenerate — this works in every MCP client, including ones that do not pass tool-result images to you. Costs one small vision chat call (well under $0.01 on the default model).
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"image_url": "<image_url>",
"brief": "<brief>"
},
"name": "review_image"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| brief | Yes | What the image is supposed to show — usually the prompt it was generated from, plus any requirements the user stated (exact text, aspect ratio, style). | |
| model | No | Vision-capable chat model that does the looking. gpt-5.6-luna (default, cheapest) or claude-sonnet-5 for a more careful read. | gpt-5.6-luna |
| image_url | Yes | The image to review: a URL returned by generate_image, any public https:// URL, a LOCAL FILE PATH, a localhost URL, or a data: URI. |