NemoClaw Windows Setup: WSL 2, Docker and OpenClaw

NemoClaw does not run as a native Windows application. NVIDIA’s current Windows setup runs it inside WSL 2. The documented baseline is Windows 10 build 19041 or newer (or Windows 11), WSL 2 with Ubuntu 24.04, and Docker Desktop using its WSL 2 backend. After the Windows bootstrap, run NemoClaw commands in the Ubuntu terminal—not PowerShell.

By using WSL 2 (Windows Subsystem for Linux), you can run an Ubuntu environment inside Windows and install NemoClaw there. This guide walks through WSL 2 and Docker Desktop, NVIDIA GPU passthrough, the NemoClaw installer and the first agent response. Check NVIDIA’s latest quickstart before copying commands because prerequisites can change.

We also cover the common repo errors you’ll likely hit along the way and how to fix them. All commands are single-line and copy-paste friendly — no backslashes, no multi-line pipes. If you prefer to watch instead of read, the full video walkthrough is linked below.

NemoClaw WINDOWS Setup Guide (It Actually WORKS)

Play videoWatch on YouTube

Step 1: Install WSL2 with Ubuntu

PowerShell (Admin):

wsl --install -d Ubuntu-24.04

After restart, in Ubuntu:

sudo apt update && sudo apt upgrade -y

---

Step 2: Enable systemd

sudo nano /etc/wsl.conf

Add:

[boot] systemd=true

PowerShell:

wsl --shutdown

Reopen Ubuntu, verify:

systemctl is-system-running

---

Step 3: Docker Desktop

1. Install Docker Desktop for Windows: https://www.docker.com/products/docker-desktop/
2. Settings → Resources → WSL Integration → toggle on Ubuntu → Apply & Restart

Verify in Ubuntu:

docker run hello-world

---

Step 4: NVIDIA GPU passthrough

1. Install latest Windows NVIDIA driver: https://www.nvidia.com/Download/index.aspx
2. Do NOT install a Linux NVIDIA driver inside WSL2

In Ubuntu — add signing key:

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg

Add repo (use the .list file URL, not the bare directory):

curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

Install:

sudo apt update sudo apt install -y nvidia-container-toolkit sudo nvidia-ctk runtime configure --runtime=docker

Restart Docker Desktop, then:

sudo apt install -y nvidia-cuda-toolkit

Verify (both must work):

nvidia-smi nvcc --version

If nvidia-smi fails → update Windows NVIDIA driver → wsl —shutdown → retry.

---

Step 5: Node.js 20+

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash - sudo apt install -y nodejs node -v && npm -v

---

Step 6: Install NemoClaw (CLI + onboard wizard)

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

The wizard runs automatically through [1/7] to [7/7]:

• [4/7] — Enter your NVIDIA API key from https://build.nvidia.com
• [5/7] — Choose cloud model (default: nemotron-3-super-120b-a12b)
• [7/7] — Accept suggested policy presets (pypi, npm) by pressing Y

After it finishes:

source ~/.bashrc nemoclaw --version openshell --version

---

Step 7: Connect and test

Check sandbox status

nemoclaw (sandbox name) status

Connect:

nemoclaw (sandbox name) connect

Inside the sandbox, launch chat:

openclaw tui

Or test via CLI:

openclaw agent --agent main --local -m "hello" --session-id test

Exit sandbox:

exit

Check logs if anything feels off:

nemoclaw boxplant logs --follow

---

Step 8: Harden WSL2

sudo nano /etc/wsl.conf

Full config:

[boot] systemd=true [interop] enabled=false appendWindowsPath=false [automount] enabled=false

PowerShell:

wsl --shutdown

Optional — memory limit (create %UserProfile%.wslconfig):

[wsl2] memory=12GB swap=8GB

---

Daily Use

nemoclaw (sandbox name) connect openclaw tui

---

Nuclear Reset (if things break)

openshell sandbox delete (sandbox name) openshell gateway destroy --name nemoclaw docker volume rm openshell-cluster-nemoclaw

Then rerun curl -fsSL https://nvidia.com/nemoclaw.sh | bash from Step 6.

---

Manual Workaround (only if Step 6 wizard fails with sandbox errors)

bash openshell sandbox delete my-sandbox 2>/dev/null openshell gateway destroy --name nemoclaw 2>/dev/null docker volume rm openshell-cluster-nemoclaw 2>/dev/null openshell gateway start --name nemoclaw openshell status openshell provider create --name nvidia-nim --type nvidia --credential NVIDIA_API_KEY=nvapi-YOUR_KEY_HERE openshell inference set --provider nvidia-nim --model nvidia/nemotron-3-super-120b-a12b openshell sandbox create --name my-sandbox --from openclaw openshell sandbox ssh my-sandbox openclaw onboard

When prompted for provider → select Custom Provider → enter https://inference.local/v1

If Anthropic key is set: unset ANTHROPIC_API_KEY