NVIDIA NemoClaw is an open-source reference stack that simplifies running OpenClaw always-on AI assistants more safely. It installs the NVIDIA OpenShell runtime and wraps it with guided onboarding, a hardened sandbox blueprint, routed inference, and layered security (Landlock, seccomp, network namespaces).
Hardware: 4+ vCPU, 8 GB RAM (16 GB recommended), 20 GB disk free.
Software:
On macOS, make sure Xcode CLI tools are installed:
xcode-select --installOne command to install NemoClaw and run the guided onboard wizard:
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bashThe wizard creates a sandbox, configures inference, and applies security policies. When it finishes you’ll see:
──────────────────────────────────────────────────
Sandbox my-assistant (Landlock + seccomp + netns)
Model nvidia/nemotron-3-super-120b-a12b (NVIDIA Endpoints)
──────────────────────────────────────────────────
Run: nemoclaw my-assistant connect
Status: nemoclaw my-assistant status
Logs: nemoclaw my-assistant logs --follow
──────────────────────────────────────────────────If nemoclaw is not found after install (common with nvm/fnm), reload your shell:
source ~/.zshrc # or source ~/.bashrcConnect to the sandbox:
nemoclaw my-assistant connectThen launch the terminal UI:
openclaw tuiOr send a one-off message:
openclaw agent --agent main --local -m "hello" --session-id testCheck sandbox status and tail logs:
nemoclaw my-assistant status
nemoclaw my-assistant logs --followcurl -fsSL https://raw.githubusercontent.com/NVIDIA/NemoClaw/refs/heads/main/uninstall.sh | bashUseful flags: --yes (skip prompt), --keep-openshell (keep the runtime), --delete-models (remove pulled Ollama models).
Note: NemoClaw is currently in alpha (since March 2026). Interfaces and APIs may change. Check the official docs for the latest.