Run ChatGPT on a VPS

Set up an always-on AI assistant on a rented server using ChatGPT via OpenAI.

Before you start

A VPS with Ubuntu 22.04 or 24.04 (2 vCPU / 4GB recommended, 1 vCPU / 2GB works)
Your VPS IP address and root password or SSH key
An OpenAI API key for ChatGPT

Treat API keys like passwords.

Open Terminal (Mac/Linux) or PowerShell (Windows):

ssh root@YOUR_SERVER_IP

apt update && apt -y upgrade
ufw allow OpenSSH && ufw --force enable

curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
apt-get install -y nodejs
npm i -g openclaw

openclaw onboard

Choose OpenAI and paste your API key when prompted.

openclaw gateway start

Check status anytime: openclaw gateway status

Create a bot with @BotFather, copy the token, then:

openclaw config set channels.telegram.botToken "YOUR_BOT_TOKEN"
openclaw config set channels.telegram.enabled true
openclaw gateway restart

Message your bot and it will connect to ChatGPT through OpenClaw.

Your assistant runs 24/7 on your VPS. Restart the gateway after reboots with openclaw gateway start.