MCP Server
Connect any MCP-compatible AI assistant to Stareezy UI — query tokens, scaffold components, validate configs, and search docs through a standard tool interface.
What it does
The MCP server exposes Stareezy UI as a set of structured tools that AI assistants can call directly. Instead of relying on training data or skill files, the assistant queries live token values, component APIs, and your project config — giving it accurate, always-up-to-date knowledge of the design system.
Available tools
get_tokensList all design tokens by category. Accepts an optional category filter (colors, spacing, radius, typography, shadows).
get_tokenGet the value and metadata for a single token by its dot-path, e.g. colors.celurenBlue.500.
list_componentsReturn a summary of all 31+ components with their prop signatures and import paths.
get_componentGet full props, usage examples, and platform notes for a specific component by name.
list_themesList all built-in themes (quasar, aurora, steins-gate, dark, light) with their brand color values.
validate_configValidate a stareezy.config.ts contents — check that themes, breakpoints, and shorthands are well-formed.
scaffold_componentGenerate a production-ready component scaffold using the correct token accessor pattern for a given component type.
search_docsFull-text search across all Stareezy UI documentation pages. Returns ranked results with relevant excerpts.
Installation
The server ships as a standalone npm package and runs via npx — no global install needed.
# Run directly (recommended)
npx @stareezy-ui/mcp-server
# Or install globally
npm install -g @stareezy-ui/mcp-server
stareezy-mcpClaude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"stareezy-ui": {
"command": "npx",
"args": ["@stareezy-ui/mcp-server@latest"],
"env": {}
}
}
}Kiro / Cursor / Windsurf — add to .kiro/settings/mcp.json or your editor's equivalent MCP config:
{
"mcpServers": {
"stareezy-ui": {
"command": "npx",
"args": ["@stareezy-ui/mcp-server@latest"],
"disabled": false,
"autoApprove": ["get_tokens", "list_components", "list_themes", "search_docs"]
}
}
}get_tokens, list_components, and search_docs to autoApprove — they are read-only and safe to run without confirmation prompts.Once configured, ask your assistant to use the server:
"Use the Stareezy UI MCP server to list all spacing tokens"
"Get the full props for the Button component from Stareezy UI"
"Scaffold a hero section using Stareezy UI tokens"The assistant will call the appropriate tools and return live, accurate results from the design system.
Usage examples
get_tokens→ Returns all spacing.[n] values with their pixel equivalents
scaffold_component→ Generates Card.tsx, Card.style.ts, Card.types.ts with correct token imports
get_component→ Returns all props, types, platform notes, and a usage example
validate_config→ Checks themes, breakpoints, shorthands and returns any issues
Configuration options
Pass environment variables or CLI flags to customize server behavior:
# Point at a custom stareezy.config.ts location
STAREEZY_CONFIG=./config/stareezy.config.ts npx @stareezy-ui/mcp-server
# Enable verbose tool call logging
STAREEZY_MCP_LOG=verbose npx @stareezy-ui/mcp-server
# Pin to a specific package version for reproducibility
npx @stareezy-ui/mcp-server@1.0.0| Variable | Default | Description |
|---|---|---|
STAREEZY_CONFIG | ./stareezy.config.ts | Path to your project config file |
STAREEZY_MCP_LOG | error | Log level: error | info | verbose |
STAREEZY_MCP_PORT | stdio | Transport: stdio (default) or http:<port> |