Banani

Type a description of a product and Banani returns a clickable, multi-screen prototype you can keep editing on a canvas rather than a flat image. Clicking an element prompts the AI to generate the screen a user would land on next, which is how the interactive flow gets built out.

Works from references, not just prompts

You can drop in screenshots, images or Figma links and Banani pulls out the layout and visual style, either to riff on or to reproduce the interface one-to-one. Styling is adjustable both by asking the AI for a different mood and by editing colour, typography and other tokens by hand.

Exports to Figma, code or an MCP connection

Projects and individual screens can be shared by link for feedback, and finished work exports to Figma, HTML/CSS or images. There is also MCP support so coding agents can read the designs directly. The founders, Vlad Solomakha and Vova Parkhomchuk, describe Banani as aimed at PMs, founders, engineers and teams short on design resources, as well as designers wanting the repetitive work handled.