NemoClaw doesn’t run on Home windows natively. It depends on Linux kernel options like Landlock, seccomp, and community namespaces — none of which exist on Home windows. Even skilled builders have struggled to get it working, and the official documentation solely helps Ubuntu 22.04 or later.
However there’s a approach. Through the use of WSL2 (Home windows Subsystem for Linux), you possibly can run a full Ubuntu setting inside Home windows and set up NemoClaw inside that. This information walks you thru each step — from putting in WSL2 and Docker Desktop, to organising NVIDIA GPU passthrough (the step most guides skip), to working the NemoClaw installer and getting your first agent response.
We additionally cowl the widespread repo errors you’ll possible hit alongside the way in which and methods to repair them. All instructions are single-line and copy-paste pleasant — no backslashes, no multi-line pipes. If you happen to want to observe as a substitute of learn, the total video walkthrough is linked under.
Step 1: Set up WSL2 with Ubuntu
PowerShell (Admin):
wsl --install -d Ubuntu-22.04
After restart, in Ubuntu:
sudo apt replace && sudo apt improve -y
Step 2: Allow systemd
sudo nano /and so on/wsl.conf
Add:
[boot]
systemd=true
PowerShell:
wsl --shutdown
Reopen Ubuntu, confirm:
systemctl is-system-running
Step 3: Docker Desktop
- Set up Docker Desktop for Home windows: https://www.docker.com/merchandise/docker-desktop/
- Settings → Assets → WSL Integration → toggle on Ubuntu → Apply & Restart
Confirm in Ubuntu:
docker run hello-world
Step 4: NVIDIA GPU passthrough
- Set up newest Home windows NVIDIA driver: https://www.nvidia.com/Obtain/index.aspx
- Do NOT set up 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 .record file URL, not the naked listing):
curl -s -L https://nvidia.github.io/libnvidia-container/secure/deb/nvidia-container-toolkit.record | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | sudo tee /and so on/apt/sources.record.d/nvidia-container-toolkit.record
Set up:
sudo apt replace
sudo apt set up -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
Restart Docker Desktop, then:
sudo apt set up -y nvidia-cuda-toolkit
Confirm (each should work):
nvidia-smi
nvcc --version
If nvidia-smi fails → replace Home 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 set up -y nodejs
node -v && npm -v
Step 6: Set up NemoClaw (CLI + onboard wizard)
curl -fsSL https://nvidia.com/nemoclaw.sh | bash
The wizard runs robotically by means of [1/7] to [7/7]:
- [4/7] — Enter your NVIDIA API key from https://construct.nvidia.com
- [5/7] — Select cloud mannequin (default: nemotron-3-super-120b-a12b)
- [7/7] — Settle for instructed coverage presets (pypi, npm) by urgent Y
After it finishes:
supply ~/.bashrc
nemoclaw --version
openshell --version
Step 7: Join and take a look at
Test sandbox standing
nemoclaw (sandbox title) standing
Join:
nemoclaw (sandbox title) join
Contained in the sandbox, launch chat:
openclaw tui
Or take a look at by way of CLI:
openclaw agent --agent predominant --local -m "hey" --session-id take a look at
Exit sandbox:
exit
Test logs if something feels off:
nemoclaw boxplant logs --follow
Step 8: Harden WSL2
sudo nano /and so on/wsl.conf
Full config:
[boot]
systemd=true
[interop]
enabled=false
appendWindowsPath=false
[automount]
enabled=false
PowerShell:
wsl --shutdown
Non-obligatory — reminiscence restrict (create %UserProfile%.wslconfig):
[wsl2]
reminiscence=12GB
swap=8GB
Each day Use
nemoclaw (sandbox title) join
openclaw tui
Nuclear Reset (if issues break)
openshell sandbox delete (sandbox title)
openshell gateway destroy --name nemoclaw
docker quantity rm openshell-cluster-nemoclaw
Then rerun curl -fsSL https://nvidia.com/nemoclaw.sh | bash from Step 6.
Guide Workaround (provided that Step 6 wizard fails with sandbox errors)
bashopenshell sandbox delete my-sandbox 2>/dev/null
openshell gateway destroy --name nemoclaw 2>/dev/null
docker quantity rm openshell-cluster-nemoclaw 2>/dev/null
openshell gateway begin --name nemoclaw
openshell standing
openshell supplier 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 supplier → choose Customized Supplier → enter https://inference.native/v1
If Anthropic key’s set: unset ANTHROPIC_API_KEY
NemoClaw doesn’t run on Home windows natively. It depends on Linux kernel options like Landlock, seccomp, and community namespaces — none of which exist on Home windows. Even skilled builders have struggled to get it working, and the official documentation solely helps Ubuntu 22.04 or later.
However there’s a approach. Through the use of WSL2 (Home windows Subsystem for Linux), you possibly can run a full Ubuntu setting inside Home windows and set up NemoClaw inside that. This information walks you thru each step — from putting in WSL2 and Docker Desktop, to organising NVIDIA GPU passthrough (the step most guides skip), to working the NemoClaw installer and getting your first agent response.
We additionally cowl the widespread repo errors you’ll possible hit alongside the way in which and methods to repair them. All instructions are single-line and copy-paste pleasant — no backslashes, no multi-line pipes. If you happen to want to observe as a substitute of learn, the total video walkthrough is linked under.
Step 1: Set up WSL2 with Ubuntu
PowerShell (Admin):
wsl --install -d Ubuntu-22.04
After restart, in Ubuntu:
sudo apt replace && sudo apt improve -y
Step 2: Allow systemd
sudo nano /and so on/wsl.conf
Add:
[boot]
systemd=true
PowerShell:
wsl --shutdown
Reopen Ubuntu, confirm:
systemctl is-system-running
Step 3: Docker Desktop
- Set up Docker Desktop for Home windows: https://www.docker.com/merchandise/docker-desktop/
- Settings → Assets → WSL Integration → toggle on Ubuntu → Apply & Restart
Confirm in Ubuntu:
docker run hello-world
Step 4: NVIDIA GPU passthrough
- Set up newest Home windows NVIDIA driver: https://www.nvidia.com/Obtain/index.aspx
- Do NOT set up 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 .record file URL, not the naked listing):
curl -s -L https://nvidia.github.io/libnvidia-container/secure/deb/nvidia-container-toolkit.record | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | sudo tee /and so on/apt/sources.record.d/nvidia-container-toolkit.record
Set up:
sudo apt replace
sudo apt set up -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
Restart Docker Desktop, then:
sudo apt set up -y nvidia-cuda-toolkit
Confirm (each should work):
nvidia-smi
nvcc --version
If nvidia-smi fails → replace Home 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 set up -y nodejs
node -v && npm -v
Step 6: Set up NemoClaw (CLI + onboard wizard)
curl -fsSL https://nvidia.com/nemoclaw.sh | bash
The wizard runs robotically by means of [1/7] to [7/7]:
- [4/7] — Enter your NVIDIA API key from https://construct.nvidia.com
- [5/7] — Select cloud mannequin (default: nemotron-3-super-120b-a12b)
- [7/7] — Settle for instructed coverage presets (pypi, npm) by urgent Y
After it finishes:
supply ~/.bashrc
nemoclaw --version
openshell --version
Step 7: Join and take a look at
Test sandbox standing
nemoclaw (sandbox title) standing
Join:
nemoclaw (sandbox title) join
Contained in the sandbox, launch chat:
openclaw tui
Or take a look at by way of CLI:
openclaw agent --agent predominant --local -m "hey" --session-id take a look at
Exit sandbox:
exit
Test logs if something feels off:
nemoclaw boxplant logs --follow
Step 8: Harden WSL2
sudo nano /and so on/wsl.conf
Full config:
[boot]
systemd=true
[interop]
enabled=false
appendWindowsPath=false
[automount]
enabled=false
PowerShell:
wsl --shutdown
Non-obligatory — reminiscence restrict (create %UserProfile%.wslconfig):
[wsl2]
reminiscence=12GB
swap=8GB
Each day Use
nemoclaw (sandbox title) join
openclaw tui
Nuclear Reset (if issues break)
openshell sandbox delete (sandbox title)
openshell gateway destroy --name nemoclaw
docker quantity rm openshell-cluster-nemoclaw
Then rerun curl -fsSL https://nvidia.com/nemoclaw.sh | bash from Step 6.
Guide Workaround (provided that Step 6 wizard fails with sandbox errors)
bashopenshell sandbox delete my-sandbox 2>/dev/null
openshell gateway destroy --name nemoclaw 2>/dev/null
docker quantity rm openshell-cluster-nemoclaw 2>/dev/null
openshell gateway begin --name nemoclaw
openshell standing
openshell supplier 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 supplier → choose Customized Supplier → enter https://inference.native/v1
If Anthropic key’s set: unset ANTHROPIC_API_KEY
