Open WebUI: install a self-hosted chat for Ollama with Docker

Open WebUI is a self-hosted chat interface that runs on your own computer or server and talks to models through Ollama or any OpenAI-compatible API. On October 23, 2026 it installs with one Docker command or with pip, the first account you create becomes the administrator and its license is permissive with one condition: keep the Open WebUI branding if you have more than 50 users. This guide covers installation, connecting Ollama, the image variants and the settings that matter, based on the official documentation read on October 23, 2026.

What it is for

Ollama runs models, but its chat is basic. Open WebUI adds what people expect from a chat app: a conversation history, user accounts, document upload with search over your files, web search and voice. It connects out of the box to Ollama, OpenAI-compatible APIs and other providers, so the same interface can use a local model and a cloud one. It is also a progressive web app you can install on a phone or desktop from the browser.

Install it with Docker

The documentation's recommended command, with Ollama already running on the same machine, is:

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

Then open http://localhost:3000. The first start takes a minute. What each part does, according to the documentation:

• -p 3000:8080 puts the interface on port 3000; change the left number if 3000 is taken.
• -v open-webui:/app/backend/data keeps your chats, users and settings in a volume that survives updates. The documentation's advice is never to run without it.
• --add-host=host.docker.internal:host-gateway lets the container reach Ollama on your machine. Without OLLAMA_BASE_URL, the image looks for Ollama at http://host.docker.internal:11434.
• WEBUI_SECRET_KEY must be set once and kept; without a fixed key, every recreated container logs everyone out. The documentation suggests generating one with openssl rand -hex 32.

Pick the right image

Table: Open WebUI image tags (official documentation, checked October 23, 2026)

The :cuda image moves Open WebUI's own embedding, reranking and speech models to the GPU; your Ollama models use the GPU only if Ollama itself does. If they do not, our guide to Ollama not using the GPU covers the fixes.

Without Docker

Open WebUI also installs with Python. It supports Python 3.11 and 3.12, not 3.13 yet, and the documentation recommends the latest 3.11 for production:

pip install open-webui
open-webui serve

This version opens on http://localhost:8080. Set DATA_DIR to choose where the data lives.

First steps after installing

1. Create your account straight away. The first user to register becomes the administrator, and sign-up then closes automatically. If someone else on your network opens the page first, they get the admin account.
2. Keep the admin password safe. The FAQ warns that a new account created without admin access stays pending until an admin approves it.
3. Pick a model. With Ollama connected, the models you have pulled appear in the selector; our list of Ollama commands shows how to download them.
4. Add documents if you want answers from your own files, which uses retrieval-augmented generation. On the slim image this needs an external embedding provider.

New users after the first are placed in pending status and need admin approval, which is the safe default if you open Open WebUI to your team.

The license in one paragraph

Since v0.6.6 (April 19, 2025), Open WebUI uses its own license: you can use, modify and redistribute it for any purpose, but you may not remove or alter the Open WebUI branding unless you have 50 or fewer users in a 30-day period, are a substantive contributor with written permission, or have an enterprise license. Code contributed up to v0.6.5 remains under BSD-3. For a family, a small office or personal use, this changes nothing.

Open WebUI or LM Studio

If you only want to chat with a model on one computer, LM Studio or Ollama's own app is simpler; our comparison of Ollama and LM Studio explains the difference. Open WebUI makes sense when several people share one machine with a good GPU, when you want one interface for local and cloud models or when you need user accounts and permissions.