Banani vs Marmo UI

Banani is a web app that generates editable, multi-screen mobile and desktop UI designs from an idea, image, or product requirements document. Users can refine designs with text prompts or the editor, add brand assets such as logos and colors, and connect screens into interactive prototypes for presentation or user testing. Real-time collaboration is listed. Exports include Figma, HTML/CSS, and images. Plus and Pro include unlimited Figma exports as editable layers with auto layout intact, as well as MCP/Code export; the FAQ identifies the code export as HTML/CSS. MCP access is included on all plans and connects designs to AI coding agents including Claude Code, Cursor, and Codex. The Free plan includes 12 monthly and 3 daily credits. Plus costs $12 a month billed yearly, with the page also showing $20; Pro costs $30 a month billed yearly, with the page also showing $50. When credits run out, new AI generations pause until they refill or the billing cycle resets, while existing projects remain accessible. Banani was founded in 2024.

Marmo UI pairs an open-source React design system with a hosted MCP server to help AI agents compose dashboards and apps using named @marmoui/ui components. The server provides live component APIs, composition patterns and generation rules; generated files pass through a review step that checks imports, props and compositions. The library includes more than 50 production-ready components and comes with Storybook and a Figma library. Developers install the React package with npm. The free plan includes documentation, the design tools and MIT-licensed core components, and allows unlimited projects, commercial use and no rate limits. Pro adds generation guided by a DESIGN.md file with brand details, plus composition patterns, Pro blocks, a hosted design-system site and priority support. It costs $9.99 per month, and a 14-day trial is listed. Integration instructions cover Claude Code, Codex, Cursor, Gemini CLI, Windsurf and other MCP clients. Generated code belongs to the user, uses the user’s own LLM tokens and stays in the user’s repository; it does not phone home.