Why don't my Ollama models show up in Open WebUI?

Almost always the container can't reach Ollama. Confirm Ollama is running and that you used the --add-host flag on the same machine, or the correct OLLAMA_BASE_URL for a different machine.

How do I update Open WebUI?

Pull the new image and recreate the container. Chat history and accounts survive because they live in the persistent Docker volume.

Give your local AI a real chat interface: Open WebUI in one command

Q: Does Open WebUI cost anything?

No — it's free, open-source software. The whole local stack (Ollama, a model, and Open WebUI) costs nothing beyond hardware and electricity.

Q: Can multiple people use Open WebUI at once?

Yes, with separate accounts and chat histories. One GPU answers one request at a time at full speed, so simultaneous heavy use means taking turns.

By Daniel S. · July 10, 2026

If you followed the 30-minute Ollama setup, you have a working local AI — in a black command-prompt window. Open WebUI fixes that: it's a free, self-hosted web page that looks and works like ChatGPT but talks to your models on your machine. One Docker command gets it running, it adds the login screen that bare Ollama famously lacks, and everyone in the house can use the same AI from their browser.

TL;DR

Open WebUI is a self-hosted chat website for your local models: conversation history, multiple models in a dropdown, document upload, and user accounts — running entirely on your hardware.
Fastest install is one Docker command (below). No Docker? There's a Python route.
It adds real login accounts in front of your AI — a genuine security upgrade over Ollama's open, no-authentication API, if you still keep the whole stack off the public internet.
Browse to http://localhost:3000, create the admin account, pick your model, chat. ⏱️ verified 2026-07-10
Other devices on your home network can use it too — that's the sane way to get AI on your phone, not port-forwarding.

What is Open WebUI and why add it?

The direct answer: it's the missing front end. Ollama is an engine — it loads models and answers requests, but its built-in interface is a text prompt in a terminal. Open WebUI is a web application that sits in front of that engine and gives you the experience you already know from ChatGPT: a chat window with memory of your past conversations, a model picker, file upload so you can ask questions about a document, and separate accounts for separate people.

Nothing about your privacy setup changes: Open WebUI runs on the same machine (or another machine on your network) and talks to Ollama locally. Your prompts still never leave the house — the reason you went local in the first place (why that matters).

What do I need before starting?

Two things: a working Ollama install with at least one model pulled (the previous guide gets you there), and a way to run Open WebUI — Docker Desktop being the easiest on Windows and Mac. Docker is a free tool that runs applications in tidy self-contained packages; if you've never used it, install Docker Desktop from docker.com first, and the rest is one command.

The one-command install (Docker)

With Ollama already running on the same machine, open a terminal and run (Open WebUI quick start, ⏱️ verified 2026-07-10 against the project's README):

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

What the pieces mean, so it's not an incantation:

-p 3000:8080 — the page will be at port 3000 on your machine.
--add-host=host.docker.internal:host-gateway — lets the container find the Ollama server running on your PC.
-v open-webui:/app/backend/data — keeps your accounts and chat history in a persistent volume so they survive updates and restarts.
--restart always — it comes back up on its own after a reboot.

Give it a minute to start, then browse to http://localhost:3000.

If Ollama runs on a different machine on your network, add -e OLLAMA_BASE_URL=http://<that-machine's-IP>:11434 to the command instead of the --add-host flag.

What if I don't want Docker?

There's a Python path: Open WebUI installs with pip install open-webui and starts with open-webui serve (getting started docs, ⏱️ verified 2026-07-10). The project pins this to Python 3.11 specifically — not just "any recent Python" — and it puts the app directly on your system rather than in a container. It works, but Docker is the better default here: updates are cleaner, removal is total, and the project publishes its images with Ollama connectivity already wired. If the word "Python environment" means nothing to you, use Docker.

First run: the two minutes that matter

The first account created on a fresh Open WebUI becomes the administrator — so create yours immediately after install. Then:

Sign in, and check the model dropdown at the top of the chat — your Ollama models should already be listed (Open WebUI auto-detects a local Ollama).
Pick one, say hello, and confirm you get a reply.
In the admin settings, decide how new users join: open sign-up (fine for a family LAN), or admin-created accounts only (tighter).

From any other device on your Wi-Fi, browse to http://<your-PC's-IP>:3000 — same page, same login, no cloud involved.

The security paragraph (short version of a longer story)

Open WebUI genuinely improves your posture: bare Ollama answers anyone who can reach its port, with no login — that's how 175,000 home AI servers ended up open to the internet. Putting a login-protected web app in front, and letting family devices use it rather than the raw Ollama port, is the right shape.

But the rule from the security post still governs: none of this goes on the public internet. Don't port-forward 3000 any more than you would 11434. A login page on the open web invites password-guessing and exposes the app's own bugs to everyone. For access away from home, a private tunnel (Tailscale or WireGuard) remains the answer — your devices get in, the internet doesn't.

When this won't help

If you're the only user and you live in the terminal, ollama run may be all you need. The UI earns its keep with history, documents, multiple models, and multiple people.
It doesn't make models better or faster. Same engine, same speed, same answers — this is presentation and management, not capability.
Low-spec machines feel it. It's a full web application; on a very modest PC alongside a model already eating your RAM, the extra load is noticeable.
If you wanted "AI on my phone from anywhere" with zero setup, that's not this — that's the tunnel setup, a separate 20 minutes.

FAQ

Does Open WebUI cost anything? No — it's free, open-source software. The whole stack (Ollama + a model + Open WebUI) costs nothing but hardware and electricity.

Do my chats in Open WebUI stay private? Yes — conversations are stored in the app's own database on your machine and the models run locally. Nothing is sent to a cloud service.

Can multiple people use it at once? Yes, with separate accounts and separate chat histories. Be realistic about hardware though: one GPU answers one request at a time at full speed, so simultaneous heavy use means taking turns.

My models don't show up in the dropdown — what's wrong? Almost always the container can't reach Ollama. Confirm Ollama is running, and that you included the --add-host flag (same machine) or the correct OLLAMA_BASE_URL (different machine).

How do I update it? Pull the new image and recreate the container (docker pull ghcr.io/open-webui/open-webui:main, then remove and re-run). Your data survives — that's what the -v open-webui:... volume is for.

Last updated 2026-07-10. Time-sensitive: the Docker command and port defaults are current per the Open WebUI docs as of this date — check docs.openwebui.com if you're reading this months later.

Sources (verified 2026-07-10): Open WebUI quick start, Open WebUI getting started, open-webui on GitHub (README used for command verification).