Google Stitch Just Became an AI Figma (And It's Free)

Google Labs’ Stitch has shifted from simple screenshot-to-design experiments into an agentic, Figma-like workflow for generating full UI systems—complete with exportable design systems, instant prototypes, and code-ready outputs. The headline change is a new native design canvas powered by “design agents” that can build layouts from prompts and, crucially, pull styling context from existing websites. That matters because it turns design from a one-off mockup task into a repeatable pipeline: capture a brand’s look, generate a structured design system, iterate on screens, then export artifacts that plug into development tools.

At the core of the update is an agentic approach that blends capabilities associated with Gemini text models and image models. Stitch can spin up multiple design agents at once, letting users choose model tiers such as Gemini 3 flash and a Pro option. Instead of producing only visual screens, Stitch now generates a design system scaffold that includes primary and secondary color palettes, font selections, and styling details for UI elements like icons and buttons. It also introduces a “design.md” file—analogous to coding agents’ agents.md concept—that wraps a design-system toolkit. That file can be edited graphically in a theme editor and also exported as text for use in code editors or other workflows, making it easier to standardize brand guidelines across many projects.

One of the most practical features is the ability to pass a URL and have Stitch extract design standards from that site. Colors, fonts, and other visual cues become context for generating the new design system and design.md documentation. In the demo, a resort website in Thailand served as the source, and Stitch quickly produced a palette and typography that matched the reference site’s “vibe.” The tool also supports iterative page generation—such as creating separate pages for items found in a navigation bar—and then wiring those pages together so users can preview navigation and make targeted edits.

Stitch’s iteration loop extends beyond static designs. Users can generate instant prototypes, move between pages, and apply AI changes to specific elements. It can also produce multiple variations of a page based on design direction (for example, shifting toward a “more holistic natural food look”), generating placeholder imagery via an image model. For interaction, Stitch adds voice-driven “vibe design,” using a Gemini Live bidirectional model so users can talk to the interface while it updates the design.

Export options tie the design work directly into building. Stitch can export to AI Studio, generating code such as an X.js JS app and adding components like authentication and databases via prompts. It can also work with MCP and skills for coding-agent workflows. Other export paths include Figma and React, plus instant mockup prototypes. The workflow even generates a project brief resembling a product requirements document that includes the design system and palette. Stitch is positioned as a free, practical alternative for teams that want to move from reference sites to working UI quickly—without requiring deep design expertise—while still producing structured artifacts developers can reuse.