@pyck/mcp
v0.1.0-alpha.0
Published
MCP server exposing @pyck/react component metadata, props, and usage examples to LLM-powered tools.
Readme
@pyck/mcp
MCP server that exposes @pyck/react component metadata, props, and usage examples to LLM-powered tools.
Install
The published package ships pre-generated component data, so no extra setup is required.
Claude Code
claude mcp add pyck-mcp -- npx -y @pyck/mcpCursor / VS Code (.cursor/mcp.json or .vscode/mcp.json)
{
"servers": {
"pyck-mcp": {
"command": "npx",
"args": ["-y", "@pyck/mcp"]
}
}
}Available tools
| Tool | Description |
|------|-------------|
| pyck_get_components | List all component IDs and exports. Call this first. |
| pyck_get_component_props | Get imports, exports, and recipe variants for a component. |
| pyck_get_component_examples | List available usage examples for a component. |
| pyck_get_component_example | Get the source code of a specific example. |
Local development
When working inside this monorepo, point your client at the workspace copy instead of the published package:
{
"servers": {
"pyck-mcp": {
"command": "bun",
"args": ["run", "--cwd", "/path/to/frontend/packages/mcp", "src/bin.ts"]
}
}
}Useful scripts:
bun run generate # rebuild generated/components.json from @pyck/react source
bun run dev # regenerate data + start server (stdio)
bun run build # compile dist/bin.mjs for publish
bun run inspect # open MCP Inspector UI
bun run test # run tests
bun run lint # biome check
bun run typecheck # tscComponent data is generated at publish time via the prepack script, so the published package is always in sync with the @pyck/react source it was built against.
