Run Your AI Assistant 24/7 for $0/Month
Set up an always-on AI agent on Oracle Cloud's free tier. 4 CPU cores, 24GB RAM, 200GB storage. No credit card charges, no trial period, no surprise bills. Ever.
What You'll Build
An AI agent that runs 24/7 on a cloud server you never pay for. Not a 30-day trial. Not a "free tier that quietly starts billing you." Oracle's Always Free ARM tier gives you a real server with 4 CPU cores, 24GB of RAM, and 200GB of storage, permanently, no credit card charges.
Your agent stays on while you sleep. It checks your email at 6am, pulls your calendar at 7, and sends you a morning briefing before you've had coffee. It monitors your inbox all day. It runs your automations overnight. And the hosting bill is exactly zero.
Why This Works
Most people run OpenClaw on their laptop. Works great until you close the lid. Then your agent goes dark. No morning briefings. No background monitoring. No overnight automations. You open your laptop the next morning and your agent missed 12 hours of being useful.
The obvious fix is a cloud server, but that means $5-20/month for a VPS. Not a lot of money, but it's another subscription. Another thing to manage. Another bill.
Oracle's free tier is genuinely free. They make their money on enterprise customers spending millions. The free tier is marketing. That's it. They give you a server that would cost $30-50/month anywhere else, and they eat the cost because they're hoping you'll eventually become a big customer. You probably won't. But the server keeps running anyway.
Prerequisites
- An Oracle Cloud account (free signup at oracle.com/cloud/free)
- A Tailscale account (free at tailscale.com)
- 30 minutes and a cup of coffee
Heads up: Oracle's free tier signup can be finicky. Some regions run out of capacity. If instance creation fails, try a different availability domain or wait a few hours. The community has a signup guide for common issues.
Step 1: Create Your Server
Log into Oracle Cloud Console and go to Compute, then Instances, then Create Instance.
Configure it like this:
- Name: openclaw
- Image: Ubuntu 24.04 (aarch64)
- Shape: VM.Standard.A1.Flex (Ampere ARM)
- OCPUs: 2 (you get up to 4 for free)
- Memory: 12GB (you get up to 24GB for free)
- Boot volume: 50GB (you get up to 200GB for free)
- SSH key: Add your public key
Click Create and note the public IP address.
You're leaving half the resources unused on purpose. That gives you room to run other things later, like a local AI model, a database, or a second agent.
Step 2: Initial Setup
SSH into your new server and get it ready:
ssh ubuntu@YOUR_PUBLIC_IP
# Update everything
sudo apt update && sudo apt upgrade -y
sudo apt install -y build-essential
# Set a hostname
sudo hostnamectl set-hostname openclaw
# Keep services running after you disconnect
sudo loginctl enable-linger ubuntu
Step 3: Install Tailscale
Tailscale creates a private network between your devices. Your server gets a permanent address that works from anywhere, no need to remember IP addresses or open firewall ports.
curl -fsSL https://tailscale.com/install.sh | sh
sudo tailscale up --ssh --hostname=openclaw
From now on, connect via ssh ubuntu@openclaw from any device on your Tailscale network.
Step 4: Install OpenClaw
curl -fsSL https://openclaw.ai/install.sh | bash
source ~/.bashrc
When it asks "How do you want to hatch your bot?", select "Do this later." We'll configure it properly.
Step 5: Configure the Gateway
Lock it down so only you can access it, then expose it securely through Tailscale:
# Keep gateway private on the VM
openclaw config set gateway.bind loopback
# Require a token to access
openclaw config set gateway.auth.mode token
openclaw doctor --generate-gateway-token
# Expose via Tailscale (encrypted, only your devices)
openclaw config set gateway.tailscale.mode serve
Save the gateway token somewhere safe. You'll need it to connect your phone, laptop, or any messaging app.
Step 6: Connect Your AI Provider
You need a brain for your agent. Add your API key:
openclaw config set ai.provider anthropic
openclaw config set ai.apiKey sk-ant-your-key-here
Or use OpenAI, Gemini, or any provider you prefer. OpenClaw supports them all.
Step 7: Start It Up
openclaw gateway start
That's it. Your agent is running. Access the dashboard at https://openclaw.your-tailnet-name.ts.net from any device on your Tailscale network.
What You Get
For $0/month, you now have:
- 24/7 uptime. Your agent never sleeps, even when you close your laptop.
- 4 CPU cores and 24GB RAM. More than enough for OpenClaw plus local tools, scripts, and automations.
- 200GB storage. Room for files, databases, logs, and anything your agent creates.
- Secure access from anywhere. Tailscale means no exposed ports, no VPN configs, no firewall headaches.
- No billing surprises. Oracle's free tier is free. Period. No gotcha after 12 months.
ARM Gotcha
One thing to know: this is an ARM server, not x86. Almost everything works, but occasionally you'll hit a tool or binary that's x86-only. When that happens, check if there's an ARM build or use the build-essential package to compile from source. In practice, this rarely comes up with OpenClaw.
The Math
Running this same setup elsewhere:
| Provider | Comparable specs | Monthly cost |
|---|---|---|
| Oracle Free | 2 OCPU, 12GB RAM | $0 |
| Hetzner CX22 | 2 vCPU, 4GB RAM | ~$4 |
| DigitalOcean | 2 vCPU, 4GB RAM | $24 |
| AWS Lightsail | 2 vCPU, 4GB RAM | $20 |
Over a year, you save $48-288 compared to paid alternatives. And you get 3x the RAM.
What People Build on This
Once your agent is always on, things click. People set up morning briefings that land before their alarm goes off. Overnight inbox monitoring that triages email while they sleep. Automated social media posting on a schedule. Background web monitoring for competitor changes, stock alerts, or news tracking.
The server just sits there, running, costing nothing, doing work.