I Made My Local AI Accessible From Anywhere

Once I had Ollama and Open WebUI running on my Acer Predator, the next obvious problem was that I could only use it when I was sitting in front of that laptop.

If I was at a client meeting, at a conference, or even just working from another room with a different device, I had no access. My first instinct was to spin up a VPS, I’ve done it before through Hostinger, But I didn’t want to pay monthly fees just to relay traffic to a machine I already own. So I looked for alternatives and found Cloudflare Tunnel.

What Cloudflare Tunnel Does

Cloudflare Tunnel creates a secure, encrypted connection between your local machine and Cloudflare’s global network. You install a small agent called cloudflared on your laptop, point it at a local port, and Cloudflare gives you a public URL. Anyone, including you from a phone or another laptop, can reach your local service through that URL. No port forwarding required. No static IP needed. No monthly VPS cost.

Why I Chose This Over Other Options

There are other tools that do something similar. Ngrok is the most popular, but the free tier gives you a URL that changes every time you restart the tunnel, which gets annoying fast. Tailscale is excellent but it’s designed for private access between your own devices, it doesn’t give you a public URL you could share or access from any browser. Cloudflare Tunnel gives you a permanent public URL on the free tier. That’s the detail that made the decision easy.

The Security Consideration

One thing I want to be clear about: I expose Open WebUI (port 8080), not Ollama directly (port 11434). Open WebUI has built-in login authentication. Ollama’s raw API has no authentication at all — exposing it publicly would mean anyone who found the URL could use your models and your hardware. Don’t do that. Always tunnel through an authenticated interface.

A Real Constraint Worth Knowing

This setup works but only when my laptop is on and plugged in. Unlike a VPS that runs 24/7, my laptop needs to be awake and running Ollama for the tunnel to serve responses. I set Windows power settings to never sleep when plugged in, which handles it for my workflow. If you need 100% uptime, a VPS is still the right answer. But if you need flexible access during your working hours without ongoing infrastructure costs, this approach does the job.

What I Use It For

Being able to open a browser on my phone and chat with a model running on my own hardware — privately, without sending data to any cloud service — is genuinely useful. For reviewing documents, drafting, or thinking through governance frameworks on the go, it works well. It’s not perfect. But it’s mine, it’s private, and it’s free

WEEKLY BUILD NOTES

One documented agent build in your inbox, every week

Real systems, real code, the errors left in. No spam, unsubscribe anytime.