NVIDIA NemoClaw is a powerful open-source platform designed to run AI agents with enterprise-grade security. It serves as a specialized security layer built on top of the viral OpenClaw agent framework, providing the necessary guardrails for companies to deploy autonomous assistants safely.
While the stack is built primarily for Linux environments to take advantage of kernel-level sandboxing, Windows users can still harness its full potential by using the Windows Subsystem for Linux (WSL2).
This guide will walk you through the exact steps to get NemoClaw up and running on your Windows machine today.
What Is NemoClaw and Why Does It Matter
Before diving into the installation, it is important to understand why NemoClaw is becoming the standard for enterprise AI agents. Most autonomous agents today run with broad permissions, which creates significant security risks.
NVIDIA solves this by introducing OpenShell, a sandboxed runtime that enforces strict network policies and egress approval.
Think of OpenClaw as the engine that drives the agent’s logic, while NemoClaw is the safety system that ensures the agent doesn’t leak private data or execute harmful commands.
By running NemoClaw on your local Windows machine via WSL2, you get the best of both worlds: a secure Linux sandbox and the comfort of your existing Windows development environment.
Prerequisites for Windows
Before you start the setup, make sure your system meets these hardware and software requirements. Performance in AI agent tasks is heavily dependent on GPU acceleration, so having the right drivers is critical.
- NVIDIA GPU You will need a modern NVIDIA GPU (RTX 30-series or higher recommended) with at least 8GB of VRAM.
- Latest Drivers Ensure you have the latest NVIDIA Game Ready or Studio Drivers installed on your Windows host.
- WSL2 Enabled Your Windows version must support WSL2 (Windows 10 version 2004 or higher, or Windows 11).
Step 1 Install WSL2 and Ubuntu
NemoClaw requires a Linux-based environment to handle its sandboxing features correctly. Windows makes this easy through WSL2.
Open PowerShell as an Administrator and run the following command:
wsl --install
By default, this will install the Ubuntu distribution. If you already have WSL installed but need a fresh Ubuntu instance, you can install it via the Microsoft Store or by running wsl --install -d Ubuntu-22.04. Once the installation is complete, restart your computer to ensure all features are activated.
After restarting, open your Ubuntu terminal and update your package lists to ensure everything is current:
sudo apt update && sudo apt upgrade -y
Step 2 Enable NVIDIA GPU Support in WSL2
For NemoClaw to access your physical GPU from inside the virtualized Linux environment, you need the NVIDIA Container Toolkit. This is a vital step that allows Docker and other runtimes within WSL2 to communicate directly with your Windows GPU drivers.
Follow the official NVIDIA Container Toolkit installation guide inside your WSL2 terminal. Generally, this involves adding the NVIDIA package repositories and installing the nvidia-container-toolkit package using apt.
To verify that your GPU is reachable inside WSL2, run:
nvidia-smi
If you see your GPU details listed in the terminal, you are ready to proceed to the NemoClaw installation.
Step 3 Install NemoClaw
With your environment prepared, you can now install the NemoClaw CLI. NVIDIA provides a streamlined installation script that handles the setup of the core components and the NVIDIA Agent Toolkit.
Run the following command in your Ubuntu terminal:
curl -sL https://nemoclaw.run/install.sh | bash
This script will configure the environment variables and ensure that the NemoClaw binaries are correctly linked to your path. You may need to restart your terminal session after this step.
Step 4 Launch Your First Agent
Now you are ready to start using NemoClaw. The first thing you need to do is start the gateway, which manages the connections between your agents and the LLM providers (like NVIDIA NIM or local vLLM instances).
Launch the NemoClaw gateway with GPU support enabled:
nemoclaw gateway start --gpu
Troubleshooting GPU Detection
In some Windows environments, the standard command might have trouble detecting the WSL2 GPU passthrough. If you encounter an error stating that no GPU was found, you can use the fallback command provided by the underlying OpenShell system:
openshell gateway start --gpu
Once the gateway is running, you can connect your first OpenClaw agent. You can find pre-built agent blueprints in the NVIDIA NemoClaw GitHub repository to help you get started with coding, research, or automation tasks.
Running NemoClaw on Windows via WSL2 is currently the most reliable way to experience NVIDIA’s new agent platform without switching entirely to a Linux OS. It provides the high-level security of a kernel-level sandbox while letting you stay productive within the Windows ecosystem.
As AI agents become more autonomous, the security guardrails provided by NemoClaw will become indispensable for both developers and enterprises alike.
READ ALSO; Google Stitch Ends the Era of Traditional UI Design as Figma Market Value Craters
