Prep

Hardware

  • Pick your device.
    • I with with a T14s, running Debian 13 “Trixie”

Accounts

  • Setup your Nvidia account if you don’t have one.
  • Setup OpenAI Keys (Or other provider)
    • This will be handy as the agent swarm expands

Docker

Run the following command to uninstall all conflicting packages:

sudo apt remove $(dpkg --get-selections docker.io docker-compose docker-doc podman-docker containerd runc | cut -f1)

Install using the apt repository

# Add Docker's official GPG key:
sudo apt update
sudo apt install ca-certificates curl
sudo install -m 0755 -d /etc/apt/keyrings
sudo curl -fsSL https://download.docker.com/linux/debian/gpg -o /etc/apt/keyrings/docker.asc
sudo chmod a+r /etc/apt/keyrings/docker.asc

# Add the repository to Apt sources:
sudo tee /etc/apt/sources.list.d/docker.sources <<EOF
Types: deb
URIs: https://download.docker.com/linux/debian
Suites: $(. /etc/os-release && echo "$VERSION_CODENAME")
Components: stable
Signed-By: /etc/apt/keyrings/docker.asc
EOF

sudo apt update

OpenShell

UV https://docs.astral.sh/uv/getting-started/installation/

curl -LsSf https://astral.sh/uv/install.sh | sh

https://docs.nvidia.com/openshell/latest/get-started/quickstart.html

uv tool install -U openshell

Docker update and restart

echo '{"default-cgroupns-mode": "host"}' | sudo tee /etc/docker/daemon.json > /dev/null

systemctl restart docker

Update Docker

sudo usermod -aG docker UserName

newgrp docker

NemoClaw

https://docs.nvidia.com/nemoclaw/latest/get-started/quickstart.html

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

NemoClaw Setup Screen

Setup Screen

NemoClaw Model Provider

  • Choose 1, and setup NVIDIA
    • Enter Nvidia API key
    • Choose Nemotron
  • Enter Sandbox name
    • function of main agent my-assistant

NemoClaw Victory

http://127.0.0.1:18789/#token=…

NOTE:

If you’re planning to run this on a laptop as I have, and then access it via ssh & the web GUI; you may want to consider tmux to manager your ssh shell sessions. DevOps Toolbox offers a great tutorial Tmux From Scratch To BEAST MODE tmux Cheat Sheet

Congratulations

You now have your OpenShell security wrapper running around a sandboxed OpenClaw, with NemoClaw as the control layer between the two; this is the down and dirty guide from notes to myself on how to set this up again. There will be way more context provided if you follow a polished guide on youtube that I’m sure are to come soon. For my part, I’ll try to share what I learn by posting the rough notes and links to the best resources I find.

This was really just to start documenting my journey with NemoClaw (this will be an ongoing series) and get the post up.

I also wanted to showcase that it’s not required to run Ubuntu to have OpenShell and in turn NemoClaw.

Read the Docs

OpenShell

https://docs.nvidia.com/openshell/latest/index.html

NemoClaw

https://docs.nvidia.com/nemoclaw/latest/index.html

This Weekend & Beyond

  • I still have to finish reading all the docs related to OpenShell.
  • Document how to make templates for network policies.
  • Design the agent swarm.
    • Create or Remix agent profiles
      • super prompts turned into:
        • Soul.md
        • Identity.md
        • Memory.md
        • Skills.md
          • Setup the local MCP server and then policy control each agent’s access to the skill_vault (markdown file directory with an obsidian wrapper)
  • Before the swarm is up, I’ll also connect at least the orchestration agent to a few Discord servers.