Stitch and the Infinite Canvas

Context

Google Stitch is an AI-assisted UI design tool: you describe a screen in natural language and it generates an interface you can export to HTML/CSS. Its workspace is an infinite canvas: instead of fixed frames like a traditional editor, you generate and place screens freely on a boundless surface. It's how Stitch implements fast AI-driven visual exploration, the approach behind interface-generation tools.

Setup

1. Go to stitch.withgoogle.com and sign in with your Google account.
2. Create a new project: an empty canvas opens.
3. Type your first prompt in the generation panel; Stitch places the generated screen on the canvas.
4. Navigate with scroll and zoom (like Figma or Miro). There's no space limit.

State as of July 2026. Stitch evolves quickly; the exact location of controls may change.

Examples

A typical exploration flow on the canvas:

1. Prompt: "Landing page for a B2B finance app, minimal, mobile-first."
   → Stitch generates 3-4 variants and places them on the canvas.
 
2. You pick one, duplicate it alongside, and ask for a variation:
   "same screen but with the CTA above the fold."
 
3. Repeat: each iteration stays in view, side by side, for comparison.

The value of the infinite canvas is that you see every version at once. In a fixed-frame editor you'd jump between artboards; here the full history of the exploration is laid out in front of you.

Particularities

• Advantage: great for exploring and comparing variants without losing the thread. Exploration is visual and spatial.
• Advantage: each screen is independent — no state is carried between them, so a bad variant doesn't contaminate the rest.
• Limitation: Stitch keeps no system memory between prompts. Two separately generated screens can drift in style. For system consistency, the next step is Claude Design.
• Limitation: the canvas gets messy fast. Name and group the variants you care about and delete the noise before it buries your view.