Google Stitch

Use it when

You need hi-fi UI concepts fast: a vague brief, a rough screenshot, or an existing .fig / live site / codebase you want to remix—before you invest in full design-system craft inside Figma.

What it does well

• AI-native infinite canvas with a design agent you can steer while generations stream onto the board
• Start from prompts, images, voice, code context—or extract DESIGN.md from a .fig, website, or codebase
• In-place AI edits for local changes without regenerating the whole screen
• HTML-native canvas with motion (scripts, SVGs, shaders) and stitch screens into playable flows
• Hand off via .fig / DESIGN.md paths, Netlify publish, Bolt or Lovable export, plus MCP skills toward coding agents (e.g. Antigravity)

Skip it when

• Your team already treats Figma as the only source of truth and only needs light AI inside that file—prefer Figma AI
• You need a permanent component library / token system as the long-term home (Stitch is Labs exploration + bridging, not a Figma replacement)
• You expect unlimited free generations; Stitch uses daily credits—check Settings for your current cap
• You only want prompt→production app logic with no design-canvas step—Bolt/Lovable may be a more direct path (Stitch can export there, but it is still a design-first loop)

Fits your workflow

Primary loop is on the Stitch canvas (not inside Figma). Upstream: blank prompt, voice, screenshots, or existing .fig / site / code → DESIGN.md. Downstream: continue in Figma, publish a claimable Netlify URL, export to Bolt or Lovable, or use MCP skills to sync with a coding agent. Treat outputs as fast direction—finalize systems in your main design tool.