GreenVPN

OpenClaw on Oracle Cloud Free Tier: Always-On AI for $0/Month

February 26, 2026 · AI Tools

Oracle Cloud's Always Free tier is one of the best-kept secrets for running a personal AI assistant in the cloud. You get 4 ARM (Ampere) cores and 24GB RAM—enough to run OpenClaw 24/7 with headroom for Ollama or other services. No credit card charges, no expiration as long as your account stays active. This guide walks you through creating the VM, installing OpenClaw, and securing remote access with Tailscale so you can chat with your lobster from anywhere.

Why choose Oracle over paid VPS providers like DigitalOcean or Linode? The Always Free tier is genuinely free—Oracle does not charge for the Ampere A1 Flex instances, block storage, or outbound bandwidth within limits. You can run OpenClaw indefinitely at zero infrastructure cost. The only recurring expense is your AI API usage (Claude, OpenAI) if you use cloud models, or nothing if you run Ollama locally. The 24GB RAM is particularly generous; most free tiers offer 1–2GB. With 24GB, you can run Ollama with a 7B or even 13B model alongside OpenClaw, giving you a fully self-hosted AI stack with no API bills.

1 Create Your Oracle Cloud Account

Go to oracle.com/cloud/free and sign up. Oracle requires a credit card for verification but does not charge for Always Free resources. Choose your region carefully—once set, changing it is difficult. US-Ashburn, EU-Frankfurt, and APAC-Tokyo are popular. After verification, you will land in the Oracle Cloud Console.

During signup, Oracle may run a small authorization charge (typically $1) that is refunded within a few days. This is standard for cloud providers to prevent abuse. Ensure your card supports international transactions. If you have had an Oracle account before, note that each person can only have one Always Free tier per lifetime—creating a new account after closing one may not grant free resources again. Pick a region close to you or your target users for lower latency to AI APIs and messaging services.

2 Launch an Ampere VM

Navigate to ComputeInstancesCreate Instance. Name it openclaw. Under "Shape", select VM.Standard.A1.Flex. Allocate 4 OCPUs and 24GB RAM (the maximum for the free tier). Choose Ubuntu 22.04 or Oracle Linux as the image. Generate an SSH key pair and download the private key—you will need it to connect. Create the instance and wait 3–5 minutes for provisioning. Note the public IP address.

The A1 Flex shape uses ARM64 (Ampere) processors. OpenClaw and Node.js support ARM natively, so you will have no compatibility issues. If you plan to run Ollama, ensure you pull ARM-compatible models (most are). For boot volume, the default 50GB is sufficient; OpenClaw uses under 1GB, and Ollama models can be 4–8GB each. You can add a block volume later if needed. Save your SSH private key in a secure location—you will use it every time you connect. On first login, Ubuntu may prompt you to change the default user password; follow the prompts.

3 Configure the Firewall

Oracle uses a default security list that blocks most inbound traffic. Add an ingress rule for SSH (port 22) from your IP, and optionally for port 18789 if you plan to access the OpenClaw Control UI directly. For better security, use Tailscale and avoid exposing 18789 to the internet. In the VCN → Security Lists → Default, add rules for TCP 22 (SSH) and TCP 18789 (OpenClaw) with source 0.0.0.0/0 if you need public access, or restrict to your IP.

Oracle's default security list is attached to your VCN's subnet. If you created a new VCN during instance creation, ensure the instance's subnet uses that security list. For SSH, restricting the source to your IP (e.g., 203.0.113.50/32) reduces exposure to brute-force attacks. For production, consider installing fail2ban to block repeated failed login attempts. If you use Tailscale, you do not need to open 18789 publicly—Tailscale creates a mesh network and you access the VM via its Tailscale IP.

4 Install OpenClaw

SSH into your instance: ssh -i your-key.pem ubuntu@YOUR_PUBLIC_IP. Update the system and install Node.js 22+:

sudo apt update && sudo apt upgrade -y
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt install -y nodejs
curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard --install-daemon

The one-line installer handles Node.js if needed. Run the onboarding wizard to configure your AI model (Claude, OpenAI, or Ollama) and add Telegram, Discord, or WhatsApp. Use --install-daemon to install a systemd service so OpenClaw starts on boot.

Have your API keys ready before starting the wizard. For Claude, get a key from console.anthropic.com; for OpenAI, from platform.openai.com. If you prefer local models, install Ollama first with curl -fsSL https://ollama.com/install.sh | sh, then pull a model like ollama pull llama3.2. During onboarding, select "Ollama" and use http://localhost:11434. For Telegram, create a bot with @BotFather, copy the token, and paste it when prompted. Test the connection by sending a message to your bot—you should get a response. If the VM cannot reach AI APIs (e.g., due to regional restrictions), configure a VPN or proxy on the server; GreenVPN's 1000Mbps bandwidth ensures fast, reliable API access.

5 Secure Access with Tailscale

Tailscale gives you a private IP (e.g., 100.x.x.x) to reach your VM without exposing SSH or OpenClaw to the internet. Install Tailscale on the VM and on your phone or laptop, then connect to http://100.x.x.x:18789 for the Control UI. This is the recommended approach from the OpenClaw docs for VPS deployments.

To install Tailscale on Ubuntu: curl -fsSL https://tailscale.com/install.sh | sh, then sudo tailscale up. Authenticate with your Tailscale account; the VM will get a 100.x.x.x address. Install Tailscale on your phone or laptop from the app store, sign in with the same account, and you can reach your OpenClaw at http://100.x.x.x:18789 from anywhere. No port forwarding, no dynamic DNS—Tailscale handles it. For additional security, enable fail2ban: sudo apt install fail2ban and configure it to protect SSH. Consider automatic security updates: sudo apt install unattended-upgrades and enable it for security patches.

Backup and Maintenance

Back up your ~/.openclaw directory regularly. Use rclone to sync to Google Drive, S3, or another cloud storage. A simple cron job can run daily: 0 2 * * * tar -czf /home/ubuntu/backups/openclaw-$(date +\%Y-\%m-\%d).tar.gz ~/.openclaw. Enable automatic security updates with unattended-upgrades. Monitor your Oracle billing dashboard to ensure you stay within Always Free limits. If your instance is terminated (e.g., for inactivity), you can restore from backup to a new VM in minutes. Keep your SSH key and Tailscale setup documented so you can reconnect quickly. The Oracle free tier has no time limit—as long as your account is active and you use only Always Free resources, you pay nothing. This makes it an excellent choice for students, hobbyists, and developers who want a personal AI assistant without infrastructure costs. Some users run multiple services on the same VM—OpenClaw, Ollama, a personal wiki, or a small web app. With 24GB RAM, you have room to experiment. Just monitor resource usage with htop or free -h to ensure you do not run out of memory. The ARM architecture is efficient; many report that the Oracle free tier performs well for typical OpenClaw workloads.

GreenVPN: Reliable Access to Your Cloud AI

Your Oracle Cloud OpenClaw needs to reach Claude, OpenAI, and messaging APIs. In regions where these are throttled or restricted, GreenVPN ensures your AI stays responsive. With 1000Mbps gigabit bandwidth, 70+ global server locations, and 10+ years of stable operation, GreenVPN keeps your cloud-based AI assistant fast and reliable.

  • ✅ 1000Mbps — no lag for API calls or Telegram
  • ✅ 70+ countries — low latency to AI providers worldwide
  • ✅ From $1.5/month — affordable for developers
  • ✅ 30-day money-back guarantee — try risk-free
  • ✅ 10+ years stable — trusted globally
Start Free Trial — From $1.5/mo

FAQ

Will Oracle really not charge me?

Always Free resources (Ampere A1 Flex, block storage, bandwidth) do not incur charges as long as you stay within limits. Avoid upgrading to paid shapes or adding extra storage. Monitor your billing dashboard to confirm zero usage.

Can I run Ollama on the same VM?

Yes. With 24GB RAM, you can run Ollama alongside OpenClaw. Install Ollama, pull a model like Llama 3.2 or Mistral, and point OpenClaw to http://localhost:11434 during onboarding. This gives you a fully free, self-hosted AI stack.

What if my Oracle VM is terminated?

Oracle may reclaim idle resources. To avoid this, ensure your instance shows activity—running OpenClaw 24/7 with periodic Telegram or API calls usually suffices. Back up ~/.openclaw regularly to another cloud or local machine so you can restore quickly if needed.

How do I reduce API costs when using Claude or OpenAI?

Use Ollama for simple tasks and reserve Claude/GPT for complex ones. OpenClaw supports model routing—configure both and let the agent choose. Alternatively, use Groq's free tier for fast inference with open models. A VPN with 1000Mbps bandwidth keeps API latency low so you are not waiting on network bottlenecks.

70+ Global Nodes · 10 Years Stable
Try GreenVPN Free