<h2>What is Stitch?</h2>
<p>Google Labs has updated Stitch, its experimental AI design tool powered by Gemini models, into what it’s calling a full “vibe design” platform. The idea: describe an app, upload a wireframe, or speak directly to the canvas, and Stitch generates responsive layouts, multiple design variations, and exportable front-end code.</p>
<p>It’s free to use, including token usage, and is aimed squarely at people who want to prototype quickly without being professional designers.</p>
<h2>What’s new in the update</h2>
<p>The original Stitch could generate UI designs from text prompts and export HTML/CSS or push to Figma. The updated version goes considerably further:</p>
<ul>
<li><strong>Infinite canvas</strong> for multi-screen projects, so the design agent can reason across an entire app rather than a single screen</li>
<li><strong>Design agent</strong> that tracks a project’s evolution and maintains visual consistency across screens</li>
<li><strong>Agent manager</strong> for running parallel design explorations simultaneously</li>
<li><strong><a href="http://design.md" target="_blank" rel="noopener nofollow">design.md</a> file</strong> — a portable, human-readable file that captures your design system rules (colors, typography, spacing) and can be passed to other tools</li>
<li><strong>URL ingestion</strong> to pull color and font styles from an existing website</li>
<li><strong>Interactive prototyping</strong> that auto-generates logical next screens based on what you’ve already built</li>
<li><strong>Voice input via Gemini Live</strong> — you can speak to the canvas in real time to request changes</li>
<li><strong>Export to Figma, React, or AI Studio</strong></li>
</ul>
<h2>The developer handoff workflow</h2>
<p>The most interesting part of this update is the MCP server SDK, which lets Stitch integrate directly with coding assistants like Claude Code, Cursor, and Gemini CLI.</p>
<p>The workflow looks like this: design your app in Stitch, export the design along with the <a href="http://design.md" target="_blank" rel="noopener nofollow">design.md</a> rules file, then hand that off to an AI coding tool to generate a React/Next.js/Tailwind implementation. Tools like Verdant can orchestrate parallel agent builds from a shared design language, so multiple agents work from the same design system simultaneously.</p>
<p>This is the “vibe coding meets vibe design” pitch — you go from a spoken description to a working prototype without writing much of anything by hand.</p>
<h2>How it competes</h2>
<p>Stitch is positioning itself against Figma in the AI-native design space, though it’s probably more accurate to say it’s targeting a different user: someone who wants to get from idea to prototype fast and doesn’t need the full feature set of a professional design tool. The free pricing and voice interface lower the barrier considerably.</p>
<p>Whether the designs hold up under real-world scrutiny — or whether the generated code is actually usable without heavy cleanup — is still an open question. But as a rapid prototyping tool, especially paired with AI coding assistants, it’s worth trying.</p>


Collections

Google Labs has updated Stitch, its Gemini-powered AI design tool, into a full 'vibe design' platform. New features include an infinite canvas for multi-screen projects, a design agent that maintains visual consistency, voice input via Gemini Live, interactive prototyping, URL ingestion for style extraction, and a portable design.md file for capturing design system rules. A new MCP server SDK enables direct integration with AI coding tools like Claude Code, Cursor, and Gemini CLI, enabling a workflow where spoken descriptions flow into working prototypes via AI-generated React/Next.js/Tailwind code. Stitch is free to use and targets rapid prototypers rather than professional designers, positioning itself as an AI-native alternative to Figma for quick ideation.

Google's Stitch brings AI-generated UI design and code from text, image, or voice prompts