Ollama already runs local models on the Mac. Open WebUI sits in front of that stack as a browser chat app. You get threads, an admin panel, and a model picker without leaving Safari or Chrome. Weights stay on the machine when Ollama serves them.
This note assumes Ollama is installed and at least one model is pulled. If you still need that first step, start with our Ollama on Apple silicon walkthrough. Pointing other apps at the raw API was covered separately in the Ollama OpenAI-compatible API tip. Here the goal is a full web UI.
Keep Ollama native on the Mac
On Apple silicon, run Ollama as the normal Mac app or CLI. That path uses Metal. Official Open WebUI docs recommend Docker for the UI, and they pick up a host Ollama at http://host.docker.internal:11434 when the container can reach the host.
Open the Ollama app so the server is listening. Confirm a model is ready:
ollama list
If the list is empty, pull a small instruct model first, then leave Ollama running.
Install Docker Desktop, then start Open WebUI
Install Docker Desktop for Mac from Docker’s site and finish its first-run setup. Open WebUI’s quick start lists Docker as the supported path on macOS.
Generate a secret key once and keep it. Without a fixed key, every recreated container logs everyone out:
openssl rand -hex 32
Start the container. Official docs map host port 3000 to the UI and mount a named volume for chats and settings:
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 hex string from openssl. The --add-host line is what lets the container talk to Ollama on the Mac.
Open localhost:3000 and create the admin account
Wait until Docker shows the container healthy. First launch can take a minute while the image finishes starting. Then open http://localhost:3000 in the browser.
The first screen offers Create Admin Account. That first user is the instance admin. Sign-up for other people stays off until you turn it on later under Admin settings. Keep the password somewhere safe.
Confirm the Ollama connection and pick a model
With no extra environment variables, Open WebUI looks for Ollama at http://host.docker.internal:11434. Check it under your avatar, then Settings, Admin, Connections. Manage Ollama API Connections should list that host URL.
If the model list is empty, Ollama may be bound only to localhost in a way the container cannot reach. Open WebUI’s connection notes say the host Ollama needs to accept that path; fix the URL in Connections or set OLLAMA_BASE_URL when you recreate the container. For a pure API client instead of a web UI, the earlier LM Studio OpenAI-compatible API tip is the other common Mac route.
Start a New Chat. Choose a model in the selector. Type a short prompt and send. Drop a file into the chat if you want to ask about a local document.
Python path without Docker
If you prefer not to run Docker, Open WebUI also ships as a Python package. Official docs support Python 3.11 or 3.12:
pip install open-webui open-webui serve
That form listens on http://localhost:8080 by default and looks for Ollama at http://localhost:11434. Set DATA_DIR if you want chats stored in a known folder. For GUI runners without a browser stack, LM Studio on Apple silicon remains the desktop alternative.
Photo: Apple
Source: Open WebUI Quick Start (docs.openwebui.com); Ollama Mac docs; Apple Newsroom (MacBook Pro lifestyle business, macOS 27 Safari tabs, macOS 27 Visual Intelligence)
