How to share one local AI with your family or office

One computer with enough memory can run a private AI for everyone at home or at work. Ollama, the engine most people use, can serve the whole network, but it has no logins, no users and no limits. This guide explains what that means, the do-it-yourself fixes, and the simple way to give each person safe access.

Updated November 3, 2026 · 7 min read

Why share one computer?

• Better answers for everyone. Good models need lots of memory. One computer with 32 GB can run a model that no phone and few laptops can.
• One place to look after. Download and update models once, instead of on every device.
• Your data stays in the building. Chats go from each phone to your own computer and back, not to a cloud AI. For a business, that also means customer data doesn’t leave your premises (see our GDPR guide).
• No per-person subscriptions. The model runs on hardware you already own.

Not sure what your computer can run? Start with the best local model for your computer’s memory, or pick your exact Mac or PC in the model finder.

What you get with plain Ollama

By default, Ollama only answers the computer it runs on. To share it, you set OLLAMA_HOST to 0.0.0.0 (or turn on Expose Ollama to the network) and give everyone the computer’s address, for example http://192.168.1.20:11434. Our phone guide has the steps. It works, but:

• No login. Anyone on the same Wi-Fi can use it, including guests. There are no users or passwords.
• Anyone can change it. The same open connection that runs chats can also download or delete models.
• No limits per person. You can’t give a child only one model, or pause someone who uses it too much.
• No view of who uses it. There’s no list of devices and no usage per person.
• No visible line. When several people ask at once, answers slow down or wait without saying why, and nobody can see where they are in line.
• Addresses change. When the router gives the computer a new address, every device stops working until you update it.

One warning: never forward Ollama’s port on your router to make it reachable from outside. Security researchers keep finding thousands of Ollama servers left open on the internet, free for anyone to use.

The do-it-yourself fixes

• A password in front of Ollama. Put a reverse proxy such as Caddy or nginx in front of it with a password, and keep Ollama itself local-only. Everyone shares one password, so you still can’t tell people apart or limit them.
• Open WebUI. A popular free web interface for Ollama with user accounts and roles. It’s a good choice if your family or team is happy to chat in a browser and you’re comfortable installing it (usually with Docker) and keeping it updated.

Both work well for technical people. For everyone else, there’s an easier way.

The easy way: PetekMind

PetekMind Hub runs on the shared Mac or Windows PC. It installs and runs Ollama for you, and becomes the only way in. Everyone else uses the PetekMind app on their Android phone or iPhone.

1. Install the hub on the shared computer. It sets up Ollama, suggests the best model for the computer’s memory, and keeps running in the background after you close the window.
2. Pair each phone. In the hub, open Devices → Add device and scan the QR code with the phone. The code works once and expires after 10 minutes. Or the phone asks to join and you click Allow after checking that the 4-digit code matches on both screens.
3. Say who it belongs to. Each device is linked to a person, so a colleague’s work phone and personal phone count as one person, with one set of limits.

Control who can do what

• Pause a person. Their devices stay paired but can’t chat until you turn them back on. Handy for bedtime or a break. For someone who has left the team, remove their devices instead.
• Limit models. Let a person see only some models, for example a small, fast one for the kids and the big one only for you.
• Block or remove a device. Blocking keeps a lost phone paired but locked out; removing cuts it off at once.
• Lock the engine to this computer. One switch makes sure Ollama only answers the hub, so pairing is the only way in, even for someone who knows the address.

Fair turns, built in

The hub answers one request at a time, first come, first served, so each answer runs at full speed. Anyone who has to wait sees their place in line on their phone. One person can have at most two requests waiting, so nobody can fill the line, and nothing that’s already running gets stopped to make room.

See who uses it

The Usage page shows what’s running now, and requests, tokens and speed over the last day, week or month, for everyone or for one person. The log stays on your computer and is kept for 90 days.

Tips for a shared AI computer

• Keep it awake. A computer that goes to sleep stops answering. Plug it in and turn off automatic sleep. The hub starts by itself when the computer starts.
• Use a cable if you can. A wired connection to the router makes replies more reliable than Wi-Fi.
• Pick a model for the busiest moment. If several people often ask at once, a slightly smaller, faster model keeps the line short.
• Away from home or the office? Phones reach the hub on the same Wi-Fi. Away from it, the PetekMind app can run a small model on the phone itself.

Quick comparison