How to install Open WebUI and use your LiteLLM models from the browser

Open WebUI is a free web interface, very similar to ChatGPT, that runs on your own computer. From it you can chat with any artificial intelligence model, upload documents, keep your conversations and create accounts for other people at home or on your team.

In this guide you will install Open WebUI and connect it to LiteLLM, the switchboard that brings all your models together (Ollama, LM Studio, OpenRouter...). That way you can pick any of them from a single screen.

Before you start

To follow this guide you need LiteLLM installed and running, with its master key at hand. If you don't have it yet, first follow our guide "How to install and set up LiteLLM to bring all your AI models together".

You also need:

• A computer with Mac, Windows 10/11 or Linux.
• About 5 GB free on disk.
• An Internet connection for the installation.

Which installation method to choose

Open WebUI can be installed in several ways. This guide explains two:

• With uv (recommended to start). It uses the same tool you already installed for LiteLLM. You don't need to install anything else.
• With Docker. Docker is a program that runs applications inside isolated "containers". It's the usual method on servers, but it requires installing Docker Desktop. We explain this option at the end.

Step 1. Open the terminal

• Mac: press ⌘ + Space, type "Terminal" and press Enter.
• Windows: open the Start menu, type "PowerShell" and press Enter.
• Linux: press Ctrl + Alt + T.

Check that uv is installed:

uv --version

If you see a version number, carry on. If not, install it as explained in step 2 of the LiteLLM guide.

Step 2. Install and start Open WebUI

With uv, a single command downloads Open WebUI and starts it. We also give it a fixed folder (DATA_DIR) to keep your data: conversations, users and settings.

Mac and Linux

DATA_DIR=~/.open-webui uvx --python 3.11 open-webui@latest serve

Windows

$env:DATA_DIR="$HOME\.open-webui"; uvx --python 3.11 open-webui@latest serve

The first time takes several minutes: it downloads Open WebUI, a suitable Python version and some components it uses to read documents. You'll see many lines of text in the terminal. When a message such as Application startup complete or Uvicorn running on http://0.0.0.0:8080 appears, it's ready.

Step 3. Create your administrator account

1. Open your browser and go to:

http://localhost:8080

2. You'll see a welcome screen ("Get started with Open WebUI"). Click to begin.
3. Fill in your name, email and a password, and click Create Admin Account.

The first account created is the administrator's. After that, sign-up for new accounts turns itself off: nobody else can create an account without your permission.

Step 4. Choose the interface language

The interface appears in English by default. To change it:

1. Click your name or avatar (bottom left) and choose Settings.
2. Under General, find Language and choose the one you want.

Step 5. Connect Open WebUI to LiteLLM

Before this step, make sure LiteLLM is running in another terminal window (with litellm --config config.yaml inside its folder).

1. Click your name or avatar and open the Admin Panel.
2. Go to Settings and then Connections. In some versions the path is Settings → Admin → Connections: the section has the same name.
3. Under Manage OpenAI API Connections, click the + button to add a connection.
4. Fill in the fields:

• URL: http://localhost:4000/v1
• API Key: your LiteLLM master key, the one starting with sk-.

5. Click Save.

Open WebUI asks LiteLLM for its list of models and adds them automatically. You don't have to type them one by one.

Step 6. Your first conversation

1. Click New Chat, top left.
2. At the top you'll see the model selector. Open it: the names you gave your models in LiteLLM appear, for example gemma-local, lmstudio-gemma or qwen-free.
3. Pick one, type your message and press Enter.

You can switch model at any time from the same selector.

Useful things you can do

• Ask questions about a document: drag a PDF or a text file into the message box and ask about its content.
• Compare models: in the model selector, the + button adds a second model to the same conversation. Both answer at once, so you can compare their replies.
• Create accounts for other people: from Admin Panel → Users you can add users and decide which models each one may use.
• Search your conversations: they're all kept in the sidebar and can be searched.

Starting it again another day

Open WebUI runs while its terminal window is open. To use it another day:

1. Start LiteLLM (and Ollama or LM Studio, if you use them).
2. Open a terminal and repeat the command from step 2. This time it starts in a few seconds.
3. Go to http://localhost:8080 and log in.

Your conversations and settings are still there, because they're stored in the .open-webui folder.

Alternative: install it with Docker

If you already use Docker, or want Open WebUI to start with your computer, this is the usual method.

1. Install Docker Desktop from docker.com (Mac and Windows), or Docker Engine on Linux. On Windows, Docker Desktop uses WSL 2: accept its installation if offered.
2. Open Docker Desktop and wait until it says it's running.
3. In the terminal, run this command (it's a single line):

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data -e WEBUI_SECRET_KEY=change-this-to-a-long-phrase --name open-webui --restart always ghcr.io/open-webui/open-webui:main

4. Wait a minute or two and go to http://localhost:3000.

Two important differences from the uv installation:

• The address is http://localhost:3000 instead of http://localhost:8080.
• Inside Docker, "localhost" refers to the container itself, not your computer. That's why in step 5 the LiteLLM URL must be http://host.docker.internal:4000/v1.

Common problems

• The first start takes a long time. That's normal: it's downloading components. Wait for the startup complete message.
• The browser can't open localhost:8080. Check that the terminal window is still open and shows no errors. If the port is busy, add --port 8081 at the end of the command from step 2 and go to http://localhost:8081.
• No models appear in the selector. Check that LiteLLM is running and that the connection's key is exactly your master key. In Connections, edit the connection: if there's a verify button, it tells you whether LiteLLM answers.
• Connection error with Docker. Use http://host.docker.internal:4000/v1 instead of localhost.
• A model appears but fails when answering. The problem is between LiteLLM and that model: check that Ollama or LM Studio are open, or that you haven't hit a free OpenRouter model's limit.
• On a Mac with an Apple chip, the Docker container hangs. If you have the DOCKER_DEFAULT_PLATFORM=linux/amd64 variable set, add --platform linux/arm64 to the Docker command.

Next steps

You now have your own ChatGPT with all your models. The next guides in the series show how to use those same models from the Cursor code editor and how to add macOS's new built-in AI model to Open WebUI.

Next guide: How to use Cursor with paid OpenRouter models and free LiteLLM models