Skip to main content
Back to Blog
AI/MLCloud ComputingDevOps
4 April 20263 min readUpdated 4 April 2026

Setting Up NemoClaw on a DigitalOcean Droplet in Minutes

NVIDIA's recent unveiling of NemoClaw at the GTC 2026 highlights an innovative open source stack designed to facilitate the secure operation of autonomous agents on the OpenClaw...

Setting Up NemoClaw on a DigitalOcean Droplet in Minutes

NVIDIA's recent unveiling of NemoClaw at the GTC 2026 highlights an innovative open-source stack designed to facilitate the secure operation of autonomous agents on the OpenClaw platform. Described as "the operating system for personal AI," OpenClaw's capabilities are expanded upon with NemoClaw's addition of sandboxing, security policies, and inference routing via NVIDIA's cloud. Though still in its alpha phase, which means there may be some instability, here's how you can set up NemoClaw on a DigitalOcean Droplet quickly and efficiently.

Prerequisites

To begin, ensure you have:

  • A DigitalOcean account
  • An NVIDIA account to generate an API key

Step 1 - Create a Droplet from the Marketplace

Visit the DigitalOcean Marketplace and locate the NemoClaw 1-Click Droplet. Select "Create NemoClaw Droplet," and opt for the CPU-Optimized plan with Premium Intel, featuring 32 GB of RAM and 16 CPUs. This setup is necessary due to the demands of running Docker containers, a Kubernetes cluster, and the OpenShell gateway.

Choose a nearby datacenter region, add your SSH key, and create the Droplet. Remember, this configuration costs $336/month, so it's advisable to destroy the Droplet when you're done to avoid unnecessary charges.

Step 2 - SSH into the Droplet

Once your Droplet is ready, connect via SSH:

ssh root@your_server_ip

The NemoClaw onboarding wizard will automatically start, checking prerequisites, installing the OpenShell CLI, and initiating the gateway.

Step 3 - Walk Through the Onboarding Wizard

Sandbox Name

The wizard will prompt you for a sandbox name. Press Enter to accept the default (my-assistant). This process involves creating the sandbox and building the container image, which takes a few minutes.

NVIDIA API Key

You'll need your NVIDIA API key to authenticate the inference routing. Generate this key from the NVIDIA Build dashboard and input it when prompted. The wizard will save the key and configure the inference provider.

Policy Presets

Next, you'll be asked about policy presets. These dictate which external services the agent can access. When prompted, decline the suggested presets unless you plan to extend the sandbox's capabilities later.

Once the setup is complete, you'll receive a summary of your sandbox details and available commands.

Step 4 - Connect to NemoClaw

To interact with your sandbox, connect using:

nemoclaw my-assistant connect

This command opens a shell within the sandbox, allowing you to launch the OpenClaw TUI:

openclaw tui

You can now engage with the AI agent within a secure environment.

Reconnecting After a New SSH Session

If you need to reconnect later, you might find that nemoclaw commands are unavailable. Fix this by adding nvm to your .bashrc, ensuring it loads on every login:

echo 'export NVM_DIR="$HOME/.nvm"' >> ~/.bashrc && echo '[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"' >> ~/.bashrc && echo '[ -s "$NVM_DIR/bash_completion" ] && \. "$NVM_DIR/bash_completion"' >> ~/.bashrc && source ~/.bashrc

Reconnect and launch the TUI as before to continue where you left off.

What’s Next

Initially, the sandbox has limited network access. To expand its capabilities, configure policy presets for services like Slack or GitHub. Refer to the NemoClaw documentation for additional integrations and setup guidance. Although still in its early stages, NemoClaw offers a glimpse into the future of always-on agents.