Quickstart

Follow these steps to get started with NemoClaw and your first sandboxed OpenClaw agent.

Note: NemoClaw currently requires a fresh installation of OpenClaw.

Prerequisites

• A Linux machine (Ubuntu 22.04+ recommended) or macOS with Docker/Colima.
• Docker Engine 20+ running.
• Node.js 20+ installed.

Install NemoClaw

Run the one-liner installer:

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

The script installs the NemoClaw CLI and its dependencies. If Node.js is not present, the script installs it via nvm.

Run the Onboard Wizard

Start the interactive setup wizard:

$ nemoclaw onboard

The wizard performs the following steps:

1. Prompts for your NVIDIA API key (get one from build.nvidia.com).
2. Creates an OpenShell gateway.
3. Registers the NVIDIA inference provider with the Nemotron 3 Super 120B model.
4. Builds the sandbox image and creates the sandbox.
5. Applies the baseline network policy.

Connect to the Sandbox

After onboarding completes, connect to the sandboxed agent:

$ nemoclaw my-assistant connect

This opens an interactive session with the OpenClaw agent running inside the sandbox.

Run Your First Prompt

Send a test message to verify everything works:

$ openclaw agent --agent main --local -m "Hello from my secure sandbox" --session-id test

The agent processes the prompt using the configured Nemotron model through the OpenShell gateway.

Next Steps

• Switch inference providers to use a different model or endpoint.
• Approve or deny network requests when the agent tries to reach external hosts.
• Customize the network policy to pre-approve trusted domains.
• Deploy to a remote GPU instance for always-on operation.
• Monitor sandbox activity through the OpenShell TUI.

Troubleshooting

If you run into issues during installation or onboarding, refer to the Troubleshooting guide for common error messages and resolution steps.