Open-source AI Converts Figma Frames to Semantic HTML/CSS, Adds Built-in Accessibility Checks

03 November 2026

Figma's latest crop of AI-generated design-to-code tools is raising the bar well beyond simple layout export. The open-source nCompose project, released in November 2026, demonstrates new capabilities by taking a Figma design and generating semantic, accessible production code for multiple front-end frameworks, all within a single intermediate representation. This leap in sophistication is changing how designers and developers use Figma to build digital experiences.

nCompose Brings AI Generation, Multi-framework Support, Accessibility Checks to Figma Login

Driving this evolution is a new generation of AI-powered design-to-code tools. Leading the pack is nCompose, an open-source project launched in November 2026 by nesterlabs-ai. By moving the conversion process into a visually-oriented intermediate representation and applying AI generation from there, nCompose is the first major project to tackle a full-featured Figma-to-code pipeline contextually.

The project takes a Figma URL as input and generates production-ready UI components for React, Vue, Svelte, Angular, and Solid. Each generated implementation includes appropriate CSS-styling, accessibility attributes, icons, and support for design-system variants. This is possible because nCompose uses the cross-framework Mitosis compiler, with Figma's designer input and context feeding an AI-enhanced reasoning stage to produce the final outputs.

What nCompose does, and where it surpasses existing converters, becomes clearer when compared to other recent Figma-to-code projects. The work by rhemaMarneni, for example, extracts design data from the Figma REST API to generate semantic HTML and corresponding CSS. It's clear, simple export with no framework-specific output or accessibility checks. Similarly, bernaferrari's deterministically generated HTML and JavaScript for multiple targets shows potential but remains basic code export compared to nCompose's higher-level association and publication.

How Multi-framework Semantic Output and Accessibility are Changing the Game

While it might seem like an esoteric detail, nCompose's commitment to generating valid and accessible components reflects a growing imperative in the Figma-to-code landscape. Previously, accessibility was a manual process completed post-implementation. Now, tools like nCompose are explicitly building it into the generation.

One key player is axe-core, a JavaScript library used by nCompose and others to identify and help correct accessibility issues in generated code against published standards like WCAG. When used alongside nCompose's analytic fidelity checks for BEM consistency, semantic HTML, CSS render accuracy, and more, it means the authored Figma design can direct not only the visual output and interactive hierarchy, but even the extent of conformance and accuracy in the generated component.

Because open-source projects generally require comprehensive documentation and open peer review, it's unsurprising to see nCompose launch with more sophistication than previously anticipated. The question now is whether these features will become standard going forward.

Where the Category is Heading

nCompose represents a notable new standard in Figma-to-code capability. Multi-framework semantic code, AI-driven layout inference from visual Figs, and automated accessibility compliance across implementations should not be considered niche or expected to remain open-source as they enter the mainstream. Their arrival signals broader change in the obvious question: If Figma generation can do that, what else could it accomplish?

nCompose's successors are already optimizing. Builder.io, for example, has begun discussing plans to add nCompose-style semantic reasoning to Visual Copilot alongside its existing Figma-to-code features. Figma itself is part of the conversation, as seen in its recent compiler-centric document on building accessible Figma autosuggest.

The challenge of contextual versus visual connection remains, as does the question of maintainability and consistency across frameworks. These are significant hurdles that may drive the next stage of Figma's evolution rather than just becoming required knowledge to achieve best results.

With nCompose now released, expectations are high and the impact noticeable. This story warrants close attention.