OpenClaw Linux Setup Recipe

Weekend #2 with 🦞 Here's my latest and the greatest recipe for installing OpenClaw.

First off, I experimented with “miniaturizing” the machine specs and have so far settled on this minimal VM resources:

• 1 CPU core
• 1024MB RAM
• 10GB Disk

I have been able to test and run things successfully on a VM virtualized using UTM. The VM runs minimal Ubuntu 24.04 LTS.

Once Ubuntu is fully installed, you have to enter a lot of long strings, for example, your Claude Code token or API tokens. So, I installed OpenSSH Server on the VM and ssh into from my Host laptop to set things up:

sudo apt update

# directly inputting into a non-gui linux machine is terrible!
sudo apt install -y openssh-server ufw
sudo ufw default deny incoming
sudo ufw default allow outgoing
sudo ufw allow OpenSSH
sudo ufw enable

# passwordless login
# sudo nano /etc/ssh/sshd_config


# needed for node
sudo apt install -y curl git
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash

nvm install 22
nvm use 22

# Finally, install OpenClaw and onboard
npm install -g openclaw
openclaw onboard

The onboarding is straight forward, but you have to enter large strings (API Keys, etc), so I recommend you ssh into the machine from a decent terminal like Ghostty.

My onboarding consists of:

1. Set up Anthropic or OpenAI key
2. Set up Telegram BOT and pair
3. Start Terminal UI (TUI) to start chatting and giving identity and instructions to your new assistant!

openclaw tui

With this minimal Linux installation, there's no desktop environment, so I will use the TUI for interacting in the backend

However, most of my interactions will be through Telegram on my phone anyway.

As a next step, I'm going to continuing hardening the VM setup to get prepared for deploying to cloud.

Enjoy Claw'ing!