Run NemoClaw with a Local LLM

Basic idea

NVIDIA NemoClaw is an open-source reference stack that simplifies running OpenClaw always-on assistants more safely. It installs the NVIDIA OpenShell runtime — an environment designed for executing agents with additional security — and connects them to local vLLM inference on your hardware platform. A single installer command (nemoclaw.sh) handles Node.js, OpenShell, and the NemoClaw CLI; the onboard wizard (or Express Install when offered) then creates a sandboxed agent, optional Brave Search, optional messaging channels (Telegram, Discord, or Slack), and a policy tier with network presets.

By the end of this playbook you will have a working AI agent inside an OpenShell sandbox, reachable through the Web UI or terminal TUI, with inference routed to local vLLM on your hardware platform. You can optionally add Telegram (and optional web search) — all without exposing your host filesystem or network beyond what you explicitly allow in policy.

What you'll accomplish

Install NemoClaw with one command (nemoclaw.sh), which pulls Node.js, OpenShell, and the CLI as needed
Complete Express Install (when offered) or walk through nemoclaw onboard with recommended settings
Open the Web UI to interact with the agent
Optionally enable Brave Search after onboarding
Optionally set up Telegram (or another messaging channel) — see Instructions, optional section at the bottom
Cleanup and uninstall with the documented uninstall flags when finished

What to know before starting

Required:

Basic use of the Linux terminal and SSH
Familiarity with Docker (permissions, docker run, optional docker group membership)
Awareness of the security and risk sections below
sudo access on the hardware platform (passwordless or interactive) — the installer and several troubleshooting steps require root privileges

Optional:

Basic networking and SSH between nodes (multi-node capable hardware only)

Supported hardware platforms

Use the matrix below to confirm your hardware platform, OS, memory, and whether multi-node applies. The same base NemoClaw workflow applies across supported hardware platforms. Multi-node inference requires multi-node capable hardware.

Hardware platform	OS	Memory	Multi-node capable hardware
DGX Spark	DGX OS (Linux)	128 GB Unified Memory	—
DGX Station	DGX OS (Linux)	Large HBM + Grace DRAM	✅ (dual-node fabric)

Prerequisites

Hardware requirements

Supported hardware platform — see Supported hardware platforms matrix above
Keyboard and monitor attached, or SSH access
Sufficient storage for the model download, vLLM container, and temporary download space (large Express models can require hundreds of GB)
Multi-node capable hardware: validated fabric connectivity between nodes (see the Multi-node tab)

Software requirements

Fresh install of DGX OS with latest updates (or equivalent supported OS on your hardware platform)
Docker installed and running
Network access to download the NemoClaw installer, container images, and model weights

Verify your system before starting:

Bash

head -n 2 /etc/os-release
nvidia-smi
docker info --format '{{.ServerVersion}}'

Expected: Ubuntu 24.04 (or your platform's supported OS), a detected NVIDIA GPU, Docker 28.x+.

Have ready before you begin

Item	When you need it
Telegram bot token (optional)	Create with @BotFather (`/newbot`). You can paste it during onboarding or when you run `nemoclaw <sandbox> channels add telegram` later.
Brave Search API key (optional)	From Brave Search API if you enable web search during onboarding, or to add it later by re-running onboarding with `BRAVE_API_KEY` set.
Hugging Face access token (multi-node only)	Required to download gated models when following the Multi-node tab.

Find model recipes

NemoClaw uses local vLLM for inference. Express Install selects a maintained model for your hardware platform. Browse tested vLLM launch settings on vLLM Recipes.

Hardware platform	Browse recipes
DGX Spark	recipes.vllm.ai — DGX Spark
DGX Station	recipes.vllm.ai — DGX Station

NOTE

Memory and storage determine what you can run. Express Install is supported on DGX Spark and DGX Station. On DGX Station, Express recommends NVIDIA Nemotron 3 Ultra; DeepSeek-V4-Flash is also supported. Large models need substantially more storage and load time.

Notice and disclaimers

The following sections describe safety, risks, and your responsibilities when running this demo.

Quick start safety check

Use only a clean environment. Run this demo on a fresh device or VM with no personal data, confidential information, or sensitive credentials. Keep it isolated like a sandbox.

By installing this demo, you accept responsibility for all third-party components, including reviewing their licenses, terms, and security posture. Read and accept before you install or use.

What you're getting

This experience is provided "AS IS" for demonstration purposes only — no warranties, no guarantees. This is a demo, not a production-ready solution. You will need to implement appropriate security controls for your environment and use case.

Key risks with AI agents

Data leakage — Any materials the agent accesses could be exposed, leaked, or stolen.
Malicious code execution — The agent or its connected tools could expose your system to malicious code or cyber-attacks.
Unintended actions — The agent might modify or delete files, send messages, or access services without explicit approval.
Prompt injection and manipulation — External inputs or connected content could hijack the agent's behavior in unexpected ways.

Participant acknowledgement

By participating in this demo, you acknowledge that you are solely responsible for your configuration and for any data, accounts, and tools you connect. To the maximum extent permitted by law, NVIDIA is not responsible for any loss of data, device damage, security incidents, or other harm arising from your configuration or use of NemoClaw demo materials, including OpenClaw or any connected tools or services.

Isolation layers (OpenShell)

Layer	What it protects	When it applies
Filesystem	Prevents reads/writes outside allowed paths.	Locked at sandbox creation.
Network	Blocks unauthorized outbound connections.	Hot-reloadable at runtime.
Process	Blocks privilege escalation and dangerous syscalls.	Locked at sandbox creation.
Inference	Reroutes model API calls to controlled backends.	Hot-reloadable at runtime.

Ancillary files

All required assets are handled by the NemoClaw installer. No manual cloning is needed for the single-node workflow.

Time & risk

Estimated time: 30–60 MIN for a first full pass (install, onboard, model download depending on choice and network). Large Express models can add substantial download time. Optional Brave, Telegram, and cloudflared steps add time if you do them in a second session. Multi-node model download and first start can take more than an hour.
Risk level: Medium — you are running an AI agent in a sandbox; risks are reduced by isolation but not eliminated. Use a clean environment and do not connect sensitive data or production accounts.
Last Updated: 07/27/2026
- Removed Agent-ready Models tab (model selection lives in Express Install); clarified Station Express recommends Nemotron 3 Ultra with DeepSeek-V4-Flash also supported; restored single upstream coding-agent starter prompt; Multi-node scoped to DGX Station