Ollama is a good way to run a local language model, but its usual interface is a terminal. Open WebUI is a self-hosted chat front end that sits in a browser, talks to Ollama, and gives you a more familiar chat layout, file uploads and an admin panel. This guide follows the official Open WebUI quick start for a Windows PC that already has, or will have, Ollama installed.
Forge PC has not audited this software, and we have not tested every install method. The steps below come from Open WebUI's official documentation, read on 13 November 2026. Commands can change between versions; if a flag looks different on the current docs page, trust that page.
If you have not installed Ollama yet, start with our Ollama and LM Studio guide. Open WebUI does not include any models of its own.
What you will need
Open WebUI's docs say the officially supported and recommended install for most people is Docker. On Windows, the same docs have a dedicated path:
Install WSL if you do not already have it, using Microsoft's official WSL install guide (linked in Sources).
Install Docker Desktop and choose the WSL 2 backend.
In Docker Desktop, go to Settings > Resources > WSL Integration, and enable integration with your default WSL distro.
Run all docker commands from a WSL terminal, not from PowerShell or Command Prompt, unless you know your Docker setup is meant to be used that way.
You also want Ollama for Windows running, so its API is available on http://localhost:11434. Open WebUI will reach that address from inside Docker via host.docker.internal.
Disk space and RAM still matter: the Open WebUI container is separate from the model files, which remain large. Our RAM for local AI guide covers that side of the PC.
One documentation warning to take seriously: do not expose this install to the wider internet until you have read Open WebUI's own hardening and sharing guides. A default install on your PC, used at localhost, is the scope of this article.
Step 1: Generate a secret key and start the container
Open WebUI tells you to set WEBUI_SECRET_KEY to a value you keep. Without a fixed key, recreating the container signs everyone out. The docs suggest generating one with openssl rand -hex 32. In a WSL terminal that command is usually already available.
The recommended docker run command, from the official quick start, 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
Replace your-secret-key with the key you generated. Keep the quotes out of casual screenshots, and do not commit that key to a public repo.
What the flags mean, according to the same page:
-p 3000:8080 — the UI is at port 3000 on your PC. Change the left number if 3000 is already in use.
-v open-webui:/app/backend/data — chats, users and settings live in this Docker volume. Do not omit it, or you will lose data when the container is recreated.
--add-host=host.docker.internal:host-gateway — lets the container reach Ollama on the Windows host.
--restart always — the container starts again after a reboot.
ghcr.io/open-webui/open-webui:main — the standard image. There is also a smaller :slim image, a :cuda image, and an :ollama image that bundles Ollama inside the container. For a PC that already runs Ollama on Windows, :main is the straightforward choice.
The first start can take a minute. The docs say you can watch progress with docker logs -f open-webui and wait until you see Application startup complete. Then open http://localhost:3000 in a browser.
Step 2: Create the admin account
The first screen offers Create Admin Account. Open WebUI is explicit: that first account is the administrator, and it manages users and instance-wide settings. Everything stays in your volume. Sign-ups then switch themselves off until you turn New Sign Ups on in the admin authentication settings, and new accounts wait as Pending until you approve them.
Note the password. The docs warn that losing the admin password locks you out of instance settings, and they point to a reset page if that happens.
Step 3: Connect Ollama and pull a model
Open WebUI has no models of its own. As the admin, go to your avatar > Settings > Admin > Connections.
If Ollama is running on the same PC, the docs say it should be picked up automatically at http://host.docker.internal:11434 (that is what the --add-host flag is for) and appear under Manage Ollama API Connections. You can then download a model from a new chat's model selector by typing its name and confirming the pull.
If the model list is empty, the official troubleshooting note is that Ollama on the host has to listen in a way the container can reach. Open WebUI documents this under its "Connection to Ollama Server" help. Do not guess a random URL; use the address in the current docs if the default does not appear.
Once a model is listed, click New Chat, pick the model, and send a message. You can also drop a file into the chat and ask questions about it, according to the same quick start.
Updating, without losing your chats
Because your data is in the open-webui volume, you can replace the container without deleting chats, as long as you keep the same volume name and the same WEBUI_SECRET_KEY. The documented manual update is: stop and remove the container (docker rm -f open-webui), pull a fresh image (docker pull ghcr.io/open-webui/open-webui:main), then run the same docker run command as before.
Do not point a :dev (pre-release) container at your everyday volume. Open WebUI warns that dev builds may run database migrations a release image cannot read back, which can stop you returning to :main. If you want to try :dev, the docs say to use a separate volume and a different port.
Our verdict
On Windows, the supported way to run Open WebUI is Docker Desktop with the WSL 2 backend, then the official docker run command with a persistent volume and a fixed WEBUI_SECRET_KEY. Open http://localhost:3000, create the admin account, and let it connect to Ollama at host.docker.internal:11434. Keep the install on your own PC unless you have followed Open WebUI's hardening guidance. We have not audited the project; the steps come from its official documentation.
