Set up Open WebUI: a private ChatGPT on your own computer

The terminal is fine for testing, but for daily use you want a proper chat window. Open WebUI is the most popular one: it looks and works like ChatGPT, keeps your chat history, reads documents and images, and runs entirely on your machine.

Before you start

You need Ollama installed with at least one model, and Docker. On Mac and Windows install Docker Desktop. On Linux, install Docker Engine.

Step 1: Start Open WebUI

What that does:

• -p 3000:8080 makes it available at http://localhost:3000.
• --add-host=... lets it reach Ollama running on your computer.
• -v open-webui:... keeps your chats and settings when it updates.
• --restart always starts it again after a reboot.

Step 2: Create your account

Open http://localhost:3000. The first account you create becomes the admin. It's stored on your machine, not online.

Step 3: Pick a model and chat

Your Ollama models appear in the menu at the top. Pick one and start typing. If the list is empty, check Ollama is running with ollama list.

Good picks: Qwen3 8B for everyday chat on 8 GB, or Gemma 3 12B if you want to drop in screenshots and photos.

Things worth setting up

Documents. Drag a PDF into the chat and ask about it. For a set of files you use often, create a Knowledge collection in the Workspace section, then type # in a chat to pull it in. See chat with your documents.

A system prompt. In settings you can give every chat standing instructions, like "Answer briefly. Use metric units."

Longer context. In a model's advanced settings, raise the context length if you paste long documents. It costs memory; see how much VRAM you need.

Use it from your phone

Anyone on your home network can open http://<your-computer's-IP>:3000. To use it away from home, put it behind a private network like Tailscale rather than opening a port to the internet. Open WebUI has logins, but it's not meant to face the open internet.

Updating

Then remove the old container and run the Step 1 command again. Your chats live in the open-webui volume, so they stay.

Without Docker

If you have Python 3.11 you can install it directly:

That serves it at http://localhost:8080.