How to Set Up Open WebUI for Local AI

Open WebUI is a polished browser-based interface for Ollama that adds persistent chat history, multi-model switching, user accounts, and custom personas — all running entirely on your own machine. This guide sets it up from scratch using Docker, then covers the key features that save the most time.

What you need

• Ollama installed and running with at least one model pulled (see the Mistral guide if needed)
• Docker Desktop installed from docker.com
• A browser — Open WebUI runs at http://localhost:3000

Step 1: Start Open WebUI with Docker

Run the following command. It pulls the image, creates a named volume for persistent data, and maps the container port to localhost:3000. The --add-host flag is required on macOS and Windows so the container can reach Ollama running on your host machine.

Step 2: Create the admin account

Open http://localhost:3000 in your browser. On first visit, Open WebUI shows a sign-up form. The first account you create becomes the admin. Fill in a name, email, and password, then click Sign Up. These credentials are stored locally and never sent anywhere.

Step 3: Connect to Ollama

Open WebUI usually detects Ollama automatically. To check, click your profile icon in the top right, go to Settings > Connections. The Ollama URL should be set to http://host.docker.internal:11434. Click the refresh button next to it. If the connection succeeds, your pulled models appear in the model selector immediately.

Step 4: Switch between models in a chat

In the chat interface, click the model name at the top to open the picker. You can switch models mid-conversation or open a multi-model comparison view that sends the same message to two models side by side. This is useful for comparing output quality between a 7B and a 12B model on the same prompt.

Step 5: Create a custom persona

Go to Workspace > Models and click New Model. Give it a name, pick a base Ollama model, and write a system prompt in the System Prompt field. Save it. The persona now appears as a selectable option in the model picker alongside the raw Ollama models, with the system prompt baked in.

You are a senior software engineer doing a code review.
Your job is to find bugs, suggest improvements, and explain tradeoffs.

Rules:
- Point out the most critical issue first.
- Cite the specific line or function when you flag a problem.
- Suggest a fix, not just a complaint.
- If the code is already correct, say so in one sentence.

Result

Open WebUI is running at localhost:3000, connected to Ollama, with persistent chat history, multi-model switching, and named personas that keep your best system prompts one click away.