Summary
The image arrives inside the call rather than as a job you have to chase.
Both tools wait for generation to finish before returning, which puts a call at 15-30 seconds and removes the polling loop a queue-based image API would need. What you can ask for is correspondingly narrow: a prompt, and for image_to_image a set of image URLs. There is no size, seed, model or output-format parameter, and nothing that lists or removes what was generated earlier.
What it is
An image server in front of Nanana AI, with one call that generates from text and one that transforms images you supply.
What you get
- text_to_image generates an image from a prompt and returns the final image URL.
- image_to_image transforms existing images, addressed by URL, against a prompt and returns the final image URL.
Requirements
A Nanana AI account and its token in NANANA_API_TOKEN. image_to_image takes its inputs as image URLs rather than local files.
Setup effort
One command plus a key — npm install -g @nanana-ai/mcp-server-nano-banana, then supply credentials