Guide

How to Setup NemoClaw

Step-by-step instructions to install, configure, and run NemoClaw. Reference: NVIDIA/NemoClaw on GitHub and NVIDIA NemoClaw. Deploy secure, always-on AI assistants with a single command.

Steps

Step 1

Prerequisites

Use a system with dedicated compute for always-on agents: NVIDIA GeForce RTX™ PCs/laptops, NVIDIA RTX™ PRO workstations, or NVIDIA DGX Station™ / DGX Spark™. You’ll need a terminal or an existing AI coding agent (e.g. OpenClaw) to run the install.

Step 2

Install NemoClaw

In your terminal, run the official install script. This adds privacy and security controls to OpenClaw and installs NVIDIA OpenShell and Agent Toolkit.

curl -fsSL https://nvidia.com/nemoclaw.sh | bash

Step 3

Onboard the agent

Start the agent and complete onboarding. From the terminal:

nemoclaw onboard

Or ask any coding agent: “Help me install nvidia.com/nemoclaw”.

Step 4

Run & verify

NemoClaw evaluates your compute and can run models like NVIDIA Nemotron™ locally for privacy and cost efficiency. A privacy router can connect agents to cloud frontier models within your guardrails. Your agent is now running with policy-based security and privacy controls.

Official repository and full docs: github.com/NVIDIA/NemoClaw, NemoClaw documentation, and NVIDIA NemoClaw.