NemoClaw

Run OpenClaw safely with NVIDIA OpenShell for always-on AI assistants.

0. SSH to Your Server

Connect to your server via SSH:

ssh root@<SERVER_IP>
  • Replace <SERVER_IP> with your server’s IP address or hostname
  • Once connected, you can run Nemoclaw commands and manage sandboxes

1. Verify Installation

Check that Nemoclaw CLI is installed and available:

which nemoclaw 
  • Confirms Nemoclaw is available at /root/.local/bin/nemoclaw
  • A user-local shim is created for easy CLI access

2. Onboard Your Server

Create your first sandbox (containerized environment for OpenClaw):

nemoclaw onboard 
  • Sets up a fresh OpenClaw sandbox inside Docker
  • The first run may take a few minutes to complete
  • Assigns a sandbox name for future operations

3. List All Sandboxes

View all sandboxes on your server:

nemoclaw list 
  • Displays names, status, and creation timestamps
  • Useful for managing multiple sandboxes

4. Check Sandbox Status

Verify your sandbox is running or initializing:

nemoclaw <SANDBOX_NAME> status 
  • Replace <SANDBOX_NAME> with your sandbox name
  • Status output shows running, initializing, or stopped

5. Connect to the Sandbox

Enter your sandbox shell:

nemoclaw <SANDBOX_NAME> connect 
  • Access the sandbox environment
  • Run OpenClaw commands and interact with agents

6. Approve Devices for OpenClaw

Before launching the TUI, you may need to pair your device.

List pending devices:

openclaw devices list
  • Shows Request IDs and device roles

Approve the device (if required):

openclaw devices approve <REQUEST_ID>
  • Replace <REQUEST_ID> with the ID shown in the previous step
  • This ensures your TUI can connect to the OpenClaw agent if pairing is needed

7. Start OpenClaw Terminal UI

Launch the terminal interface inside the sandbox:

openclaw tui 
  • Provides a text-based UI for interacting with OpenClaw agents
  • Supports chatting, automation, and sandbox management

Application Details