Integration
Replit Agent builds it. Windcraft keeps the look consistent.
Replit Agent will plan, scaffold, and iterate on an app for you in the cloud. Across all that iteration the visual language drifts. Windcraft takes the generated code into your repo and aligns it to one set of tokens and contracts.
Why iterative agent builds drift
Replit Agent works in its own cloud workspace, generating and revising over many steps. Each revision reasons locally about the change at hand, not about the spacing scale three screens away. The result is an app that runs but whose surfaces disagree — different grays, different radii, different padding — because nothing in the loop reads a shared design source.
How Windcraft aligns Replit Agent output
Replit Agent does not expose a local MCP server, so Windcraft does not pretend to plug into it. The honest flow: bring the agent’s code into your repo, then let Windcraft’s CLI and static analyzer check it against your tokens and contracts and flag the drift.
- Pull the Replit project into your repo (Git or export).
- Run npx windcraft check to surface off-system colors, radii, and spacing.
- Run npx windcraft check --fix to align the safe violations to your tokens.
- Run npx windcraft sync so tailwind.config.ts and tokens.css stay canonical.
Check Replit Agent output against your design system
# with the Replit project in your repo
npx windcraft check
npx windcraft syncFAQ
- Can Windcraft hook into Replit Agent while it builds?
- No. Replit Agent runs in its own cloud and does not expose a local MCP server, so there is nothing to connect to mid-build. Windcraft aligns the code after you bring it into your repo.
- Does this slow down the agent?
- Not at all — Windcraft is a separate step you run on your machine. The agent builds at full speed; Windcraft reconciles the result with your design system afterward.
- Where does my code live?
- On your machine. Windcraft analyzes locally and only syncs metadata — tokens, contracts, and violations — to the cloud. Your source never leaves.