MCP

The @rslab/cli package doubles as an MCP server, so agents like Claude

Code can list and install components without shelling out to add.

Claude Code

Run this once, from anywhere:

claude mcp add rslab -- npx -y @rslab/cli mcp

That's it, ask Claude to "add the pull-search animation" and it calls

add_component directly.

Cursor

Cursor doesn't have a one-command add like Claude Code, drop this into

~/.cursor/mcp.json (global) or .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "rslab": {
      "command": "npx",

      "args": ["-y", "@rslab/cli", "mcp"]
    }
  }
}

Restart Cursor, then ask it to list or add a component.

Codex

Codex CLI reads its MCP servers from ~/.codex/config.toml: add this

table:

[mcp_servers.rslab]

command = "npx"

args = ["-y", "@rslab/cli", "mcp"]

Other MCP clients

Anything else that reads an mcp.json (Windsurf and most others) uses the

same shape as the Cursor config above.

Tools

• list_components: every entry in the registry, name/type/description.

• add_component: install one or more components (and their registry

  dependencies) into a project. Needs rslab_login first for paid entries.

• check_setup: verify a project is ready for RSLAB, NativeWind v5 runtime,

  required packages, GestureHandlerRootView, each failed check comes with

  its exact fix. Same checker as npx @rslab/cli doctor.

• rslab_login: opens a browser link to authorize, no token to copy-paste.

  Shares ~/.rslabrc with the rslab CLI, so logging in from either unlocks

  both. npx @rslab/cli login does the same thing from a terminal.

Notes

The server writes files directly to disk with the same permissions as

whatever launched it, same model as the CLI, just callable by an agent

instead of a shell command.