Integration
Continue assists in your editor. Windcraft supplies the design language.
Continue brings a configurable AI assistant into VS Code and JetBrains. Whatever model you wire behind it still guesses at your design values. Windcraft adds a local MCP server so the assistant reads your tokens and contracts directly.
Why a configurable assistant still drifts
Continue lets you choose the model and the context, but design consistency is not free with either. The model completes from its training and the snippets you attach, so without your actual token values in the loop it reaches for generic grays and radii. Swapping models does not fix it — every one of them guesses unless it is handed the values to read.
How Windcraft plugs into Continue
Continue reads MCP servers from its YAML config. Declare Windcraft and the assistant can pull your tokens, contracts, and patterns as context for any UI it generates, across whichever model you have configured.
- Run npx windcraft init in your repo to generate the config.
- Add the Windcraft server under mcpServers in your Continue config.yaml.
- Ask Continue for UI; it reads the design tools as it generates.
- Run npx windcraft check to flag any drift before you commit.
Continue MCP config (config.yaml)
mcpServers:
- name: windcraft
command: npx
args:
- windcraft
- mcpFAQ
- Does it work regardless of which model I use in Continue?
- Yes. Windcraft serves values over MCP, which is model-agnostic. Whether you run Claude, GPT, or a local model behind Continue, the assistant reads the same tokens and contracts.
- Does it work in both the VS Code and JetBrains versions?
- MCP support follows Continue’s own capability in each IDE. Where Continue can call MCP servers, it can call Windcraft; the static analyzer (npx windcraft check) backstops either way.
- Does my source code leave my machine?
- No. The MCP server and analyzer run locally. Only metadata — tokens, contracts, rule violations — is synced to the cloud.