Stitch launched at I/O 2025 as a text-to-UI generator: describe what you want, get a screen back. One year later, Google Labs reintroduces it as an agentic design platform: new role, new interface, new place where the product closes.
What changed
The core of the new Stitch is the Stitch Agent: a layer that builds, reorganizes and reflows the interface as you describe what you want — by voice or text. Google explicitly talks about vibe design: a conversational workflow where the designer steers the iterations, the agent executes them in streaming, and you watch the interface take shape in real time. You can start from scratch, from a text prompt, from an existing design file, or from a codebase already written.
Where projects end up
The other half of the announcement is the export pipeline. Once the design is ready, Stitch produces a shareable link via Google AI Studio. To take the project into production there are two official paths: export to Antigravity to wire up backend logic, or publish straight to the web through Netlify. The tool remains free.
Why it matters
For designers, the interesting part isn't automatic generation — that has existed for a while — but the fact that Stitch now closes the loop down to deploy without a manual handoff between design and development. For small teams, it's the kind of shortcut that reshapes the work. For agencies and studios, it's a signal that the boundary between prototyping tool and development environment is becoming porous. As Digitimes and Google's own blog noted, the positioning is explicitly competitive: Google is trying to enter the design-tool space with a weapon competitors don't have — verticality across its own stack (AI Studio, Antigravity, Gemini).