Installing NemoClaw

InstallingNemoClaw.md

Installing NemoClaw

1. Prerequisites

Before starting, ensure you have these components installed on your Windows host:

• WSL2: Open PowerShell as Administrator and run wsl --install. Ensure you are using a modern distribution like Ubuntu 22.04 or 24.04.
• Docker Desktop: Download and install Docker Desktop.
  • In Docker Settings, go to Resources > WSL Integration and ensure your Ubuntu distro is toggled ON.
• NVIDIA API Key: Go to build.nvidia.com, sign up for a free account, and generate an API key (it will start with nvapi-).

---

2. Prepare the WSL2 Environment

Open your Ubuntu terminal from the Start Menu and run the following updates:

sudo apt update && sudo apt upgrade -y
sudo apt install -y curl git

3. Install NemoClaw

Run the official installation script provided by NVIDIA. This script will install OpenShell (the sandbox runtime) and NemoClaw itself.

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

Note: If the script asks to run the "onboard" wizard immediately, it is often safer to cancel (Ctrl+C) and do it manually, as the automated GPU detection can sometimes fail in WSL2.

4. Manual Onboarding (Recommended for WSL2)

To ensure the sandbox connects correctly to NVIDIA's cloud inference without GPU passthrough issues:

1. Start the OpenShell Gateway:

   openshell gateway start --name nemoclaw

2. Register your NVIDIA API Key:

   openshell provider create \
     --name nvidia-nim \
     --type nvidia \
     --credential NVIDIA_API_KEY=your_nvapi_key_here

3. Create the Sandbox:

   openshell sandbox create --name my-assistant --from openclaw

4. Connect and Onboard:

   nemoclaw my-assistant connect

   Once inside the sandbox, run openclaw onboard to finish the agent configuration.

---

Summary Table: NemoClaw vs. Claude Code

Feature	NemoClaw	Claude Code (Official)
Provider	NVIDIA (Open Source)	Anthropic (Proprietary)
Main Advantage	High security/sandboxing	Seamless integration with Claude
Windows Support	Via WSL2 Only	Native (Node.js/npm)
Primary Models	Nemotron, Llama 3, etc.	Claude 3.5 Sonnet

Troubleshooting

• Docker Permissions: If you see a "permission denied" error, add your user to the docker group: sudo usermod -aG docker $USER and then restart your WSL terminal.
• Memory: Ensure your WSL2 instance has at least 8GB of RAM allocated (check your .wslconfig file if performance is slow).

Are you planning to use NemoClaw for local code editing, or are you primarily looking for the most secure way to run an AI coding agent?