Open WebUI wraps your Ollama models in a full browser interface — conversation history, model switching, file uploads, and image generation. One Docker command. Prerequisite: Ollama running.
Open localhost:3000 and you get conversation history, model switching between any Ollama model, PDF upload, and optional image generation. Looks like ChatGPT. Runs on your hardware. Data stays on your machine.
docker.com/get-startedOpen WebUI connects to Ollama at http://localhost:11434. Confirm it's running before you start.
If you get “connection refused”, start Ollama first. On Mac/Windows, launch the Ollama app from your Applications folder or system tray. On Linux: ollama serve.
Don't have Ollama yet? Install it in 5 minutes →
Docker is the container runtime Open WebUI uses to ship as a single portable package. Docker Desktop is the free GUI version that includes everything you need.
Download Docker Desktop for Mac from docker.com/get-started. Pick Apple Silicon or Intel depending on your chip. Install the .dmg, launch it, and wait for the whale icon in your menu bar to stop animating.
Or install via Homebrew:
One command. Docker pulls the Open WebUI image (~900MB) and starts a container that connects to your local Ollama. The -v open-webui:/app/backend/data flag saves your conversation history between restarts.
Container exits immediately / “Conflict”
An old Open WebUI container is already registered. Remove it first: docker rm open-webui. Then re-run the command above.
Open http://localhost:3000 in your browser. The first visit asks you to create an admin account — this is local only, no email verification needed. Pick any username and password.
After signing in:
You have a full ChatGPT-style interface connected to your local Ollama models. Your data stays on your machine. Your conversations are saved. You can upload files. Switch models. All free.
The 7-minute estimate is for the Docker image pull on a fast connection. On slow hotel Wi-Fi, the ~900MB image takes longer. Once it's downloaded, startup is under 10 seconds on any modern machine.
We tested Mac (M3), Windows 11 (RTX 3080), and Ubuntu 22.04. The 87% success rate is mostly Windows Docker Desktop issues — specifically the WSL 2 requirement. If WSL 2 isn't set up correctly, Docker won't start. The Docker Desktop installer handles this automatically, but occasionally it needs a reboot to complete.
The PDF upload feature is solid for structured documents. Scanned PDFs (images of pages, not searchable text) are hit or miss — Whisper is the better tool for audio; for scanned documents, use a dedicated OCR step first.
wsl --updatein PowerShell as Administrator, then restart your machine. If Docker still won't start, download the WSL 2 kernel update package from Microsoft's docs and run it manually.curl http://localhost:11434should print “Ollama is running”). On Mac/Windows, make sure you used the --add-host=host.docker.internal:host-gateway flag in step 3. On Linux, use --network=host.-p 3000:8080 with -p 3001:8080 and open localhost:3001.docker rm open-webui. Then re-run the step 3 command. Your data is safe — it's stored in the Docker volume, not the container.Tomorrow's best AI project. Same kind of guide. Free. No spam ever.