How to Install OpenClaw on an Nvidia Jetson Orin Nano

The internet has been bursting at the seams with news about OpenClaw, the open-source personal AI assistant with super powers. I’m not quite ready to install it on my primary MacBook Pro yet, but I have an Nvidia Jetson Orin Nano which seems like the perfect candidate for a test.

Prerequisites

• If you want to install OpenClaw on a Jetson Orin Nano, you need a Jetson Orin Nano.
• You should have it set up already and have SSH access.
• You need access to some AI model, most likely API access or some subscription (OpenAI, Github Copilot, Anthropic, you name it).

Update to the latest JetPack

SSH into your Nano and update to the latest version of JetPack (6.2.2 as of February 22, 2026). I followed this video.

Feel free to also check out the official Nvidia JetPack 6.2.2 upgrade instructions.

So in my case all I had to do:

1. Run sudo vim /etc/apt/sources.list.d/nvidia-l4t-apt-source.list and update all three occurrences of 36.4 to 36.5. Save the file and exit the editor.
2. Run sudo apt update.
3. Run sudo apt dist-upgrade. This will take about 30 minutes or so. It will ask if you want to overwrite some config files. I looked at the diff for each using D, couldn’t see any differences I wanted to preserve and then pressed Y to overwrite with the new version.
4. Run sudo reboot to reboot the machine. That’s needed to actually upgrade the firmware.

Install Node

The OpenClaw installer script claims to install everything (including Node), but in my case it didn’t. So I had to install Node 24 manually using NodeSource. I simply ran these two commands:

curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash -
sudo apt install -y nodejs

Verify by running node --version. It should print 24.13.1 (or newer).

Install Linuxbrew

Since OpenClaw uses brew to install dependencies, it’s a good idea to already set up Linuxbrew before we start the OpenClaw installation.

Simply follow these steps:

sudo apt install -y build-essential curl file git

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

echo 'eval "$(/home/linuxbrew/.linuxbrew/bin/brew shellenv bash)"' >> ~/.bashrc
eval "$(/home/linuxbrew/.linuxbrew/bin/brew shellenv bash)"

Make sure it works: brew doctor.

Install OpenClaw

I tried running the installer again, and it finished successfully but the openclaw command was not available and also didn’t exist in ~/.npm-global/bin. I saw it spat out inappropriate ioctl for device during the install.

I found this GitHub issue about the problem and then tried the suggested solution:

curl -fsSL https://openclaw.ai/install.sh | bash -s -- --no-gum

I then selected the Quickstart option and Copilot as my model/auth provider. I followed the GitHub Copilot auth-flow and then selected github-copilot/claude-opus-4.6.

For channel I selected “Telegram (Bot API)” and followed the instructions to set up my lobster bot.

Next I selected “No” for “Configure skills now?”. At this point things might look a bit different for you. At first run I had selected “Yes” for configuring the skills and had selected a few. Installation failed because I hadn’t yet installed Linuxbrew. So I couldn’t finish the initial onboarding all the way through.

After sending the first “Hello” to my OpenClaw installation, I saw that it worked:

So it was pretty easy to set up. It is currently installing whisper.cpp on my behalf so it can understand voice messages I send it. I’m curious to see how performant it will be on a Jetson Nano and how useful. Running it on a Mac enables a lot of extra capabilities. That’s why quite a few people are buying dedicated Mac minis just to run OpenClaw. One of the next things I want to set up is headless Chrome with Playwright in order to use OpenClaw’s web browsing capabilities. I’ll update this as I learn more.