@app-cn/mcp
v0.1.2
Published
Model Context Protocol server for appCN. Lets AI coding agents discover, install, and correctly use appCN React Native components.
Maintainers
Readme
@app-cn/mcp
The appCN MCP server — gives AI coding agents (Claude Code, Cursor, Windsurf, Codex, …) first-class access to the appCN React Native component library. Agents can discover components, read their full docs (anatomy, the delight detail, props, usage examples, a11y notes), get the exact install command, and learn appCN's motion/haptic design tokens — so the code they write matches the system instead of guessing.
It reads from the live appCN registry (https://appcn.vercel.app/r), so it always
reflects the latest published components — nothing to update on your side.
Install
Claude Code
claude mcp add appcn -- npx -y @app-cn/mcpCursor / Windsurf (mcp.json)
{
"mcpServers": {
"appcn": {
"command": "npx",
"args": ["-y", "@app-cn/mcp"]
}
}
}VS Code (.vscode/mcp.json)
{
"servers": {
"appcn": {
"command": "npx",
"args": ["-y", "@app-cn/mcp"]
}
}
}Tools
| Tool | What it does |
| :-- | :-- |
| list_components | List every appCN component (optionally filtered to base or ai), with its one-line delight detail. |
| search_components | Find components by intent — e.g. "chat input", "voice indicator". |
| get_component | Full payload for one component: docs, source files, npm + registry dependencies, and the install command. |
| get_install_command | The exact command to add a component (appCN CLI, shadcn URL, or namespaced), per package manager. |
| get_design_guide | appCN's taste layer — motion/haptic tokens and the house rules — so generated code feels like appCN. |
Configuration
| Env var | Default | Purpose |
| :-- | :-- | :-- |
| APPCN_REGISTRY_URL | https://appcn.vercel.app/r | Point the server at a different (e.g. local) registry. |
How it relates to the shadcn MCP
appCN is a standard shadcn-compatible registry, so shadcn's own MCP can install
appCN components too (register "@app-cn" in your components.json). What this
server adds on top is the appCN-specific knowledge shadcn's generic registry JSON
doesn't carry: the delight detail, usage examples, a11y notes, and the motion/haptic
design guide — the things that make agent-authored code feel like appCN.
License
MIT
