NVIDIA Elements

@nvidia-elements/cli

NVIDIA Design System and UI Agent Harness for AI/ML Factories, Robotics, and Autonomous Vehicles.

The @nvidia-elements/cli is a dual-mode command-line tool for the Elements Design System. It provides interactive CLI commands and a Model Context Protocol (MCP) server for AI assistant integration.

Purpose

This package serves two primary modes:

1. Interactive CLI (`nve` command) - Command-line interface with interactive prompts for:

• Component API discovery and search
• Example template browsing and searching
• Template linting and validation
• Project scaffolding and health checks
• Changelog and version information
• Design token access
• Icon name lookup
• Bundled agent skill access

1. MCP Server (`nve mcp` command) - Model Context Protocol server that:

• Exposes MCP-supported tools to AI assistants (Claude, Cursor, etc.)
• Provides context-specific prompts for common tasks
• Enables AI-assisted development with Elements components
• Integrates Elements knowledge directly into AI workflows

Getting Started

The best way to get started is to run the install script.

curl -fsSL https://nvidia.github.io/elements/install.sh | bash

On Windows, run the PowerShell installer.

irm https://nvidia.github.io/elements/install.ps1 | iex

For agents and CI, invoke the canonical path directly: $HOME/.nve/bin/nve on macOS and Linux, or $env:LOCALAPPDATA\nve\bin\nve.exe on Windows.

Alternatively you can install with Node.js and npm.

npm install -g @nvidia-elements/cli

Usage

Global Options

MCP

Quick Setup

The fastest way to configure MCP is with the project.setup command:

nve project.setup

This detects your package manager, configures the MCP server for Cursor, Codex, and Claude Code, and adds Elements core dependencies to the project.

Claude Code

Install to Claude Code by adding the configuration to your .mcp.json file. The file is typically located at ~/.config/claude-code/.mcp.json or %APPDATA%\claude-code\.mcp.json on Windows.

{

"mcpServers": { "nvidia_elements": { "description": "NVIDIA Elements UI Design System (nve-*), custom element schemas, APIs and examples", "command": "nve", "args": ["mcp"] } } }

After adding the configuration, restart Claude Code for the changes to take effect. The Elements MCP tools are then available for use in your conversations.

Cursor

Install to Cursor with the MCP configuration below.

// .cursor/mcp.json

{ "mcpServers": { "nvidia_elements": { "description": "NVIDIA Elements UI Design System (nve-*), custom element schemas, APIs and examples", "command": "nve", "args": ["mcp"] } } }

Codex

Install to Codex with the MCP configuration below.

[mcp_servers.nvidia_elements]

description = "NVIDIA Elements UI Design System (nve-*), custom element schemas, APIs and examples" command = "nve" args = ["mcp"]

Prompts

NVIDIA Elements Skill

Elements provides one elements skill with reference files for artifact creation, setup checks, project integration, and migration. Install the complete directory in the current project:

nve skills.install

This writes .agents/skills/elements/ and .claude/skills/elements/. Add --global to install it for the current user at ~/.agents/skills/elements/ instead. The command replaces existing elements directories at the selected destinations so stale references do not remain.

Alternatively, install the Elements agent skill with the open skills CLI:

npx skills add https://github.com/nvidia/elements --skill elements

This route does not install the Elements CLI or configure the MCP server. Use nve project.setup for complete project setup, and continue to use the CLI or MCP tools for deterministic API lookup and template validation.

MCP Tools

Links

• Documentation
• Changelog
• GitHub Repo
• npm
• MCP Registry