Integration
Zed is fast. Windcraft keeps the speed from outrunning the design.
Zed pairs a fast native editor with an AI assistant that reads MCP context servers. Speed gets code on screen quickly; it does nothing for whether that code matches your tokens. Windcraft adds a context server so it does.
Why fast generation still drifts
Zed’s assistant generates from the open buffer and the model’s priors. The editor’s speed means more code lands faster, which means design drift also accumulates faster if nothing supplies your values. A quick edit picks a plausible gray; the next quick edit picks another. The velocity is real, but velocity without a shared source of design values just reaches inconsistency sooner.
How Windcraft plugs into Zed
Zed reads MCP servers as context servers configured in its settings. Add Windcraft and the assistant can pull your tokens, contracts, and patterns as context whenever it generates UI.
- Run npx windcraft init in your repo to generate the config.
- Add Windcraft under context_servers in your Zed settings.json.
- Zed’s assistant reads the design tools as it generates UI.
- Run npx windcraft check to flag any drift before you commit.
Zed context-server config (settings.json)
{
"context_servers": {
"windcraft": {
"command": {
"path": "npx",
"args": ["windcraft", "mcp"]
}
}
}
}FAQ
- What is a context server in Zed?
- It is Zed’s name for an MCP server the assistant can pull context from. Windcraft runs as one, so Zed reads your tokens, contracts, and patterns the same way it would any MCP source.
- Does it slow Zed down?
- No. Zed stays as fast as ever; the Windcraft server is local with a short cache and responds in well under a tenth of a second. You keep the speed and gain the consistency.
- Does my source code leave my machine?
- No. The context server and analyzer run locally. Only metadata — tokens, contracts, rule violations — is synced to the cloud.