Integration
v0 nails screen one. Windcraft keeps screens two through twenty matching.
v0 turns a prompt into a clean React screen in seconds. The trouble starts at scale: each prompt is its own little universe of spacing and color. Windcraft takes the code you paste into your repo and aligns it to one system.
Why v0 output drifts across screens
v0 generates in its own cloud from a prompt — it has no view of your repo, your tokens, or the screen you built yesterday. Every generation starts fresh, so screen two picks a slightly different gray, screen three a different card radius. The first screen is great; the twentieth is a different design system wearing the same logo.
How Windcraft aligns v0 output
Because v0 runs in its own cloud, there is no MCP server to plug in — and we will not pretend there is. The honest path: v0 writes the first draft, you bring that code into your repo, and Windcraft’s CLI and static analyzer align it to your tokens and flag what drifted.
- Generate the screen in v0 and paste the code into your repo.
- Run npx windcraft check to surface off-token colors, radii, and spacing.
- Run npx windcraft check --fix to snap the easy violations to your tokens.
- Run npx windcraft sync so your tailwind.config.ts and tokens.css stay the source of truth.
Align pasted v0 code to your design system
# after pasting v0 output into your repo
npx windcraft check
npx windcraft check --fixFAQ
- Does Windcraft connect to v0 directly?
- No. v0 generates in its own cloud and does not run a local MCP server, so there is nothing to connect to. Windcraft works on the code after you bring it into your repo — checking and aligning it to your tokens.
- So Windcraft does not change how I use v0?
- Right. Prompt v0 exactly as you do today. Windcraft is the step after: it takes the generated code and reconciles it with your design system instead of letting each screen go its own way.
- What does it actually fix?
- Off-token colors, mismatched radii, and ad-hoc spacing get flagged; the safe ones can be auto-fixed to your tokens. Anything ambiguous is reported so you decide, rather than being silently rewritten.