app·AI Tools

Open WebUI

Self-hosted web interface for chatting with local or BYOK-connected models.

FREEOPEN SOURCESELF-HOSTEDBYOK

Setup guide · 30–45 min

Visit Open WebUI →

Quick facts

Price model: Self-hosted
Starting price: Free (self-hosted)
Best for: Family or team local AI · Ollama front-end · BYOK model routing
Replaces: ChatGPT Team, Custom GPT enterprise tiers
Platforms: Self-hostedWeb
Last verified: 2026-06-22

Why it's listed

Turn a spare machine into your own AI portal—no per-message billing.

Open WebUI wraps Ollama and OpenAI-compatible APIs in a polished team-friendly UI you host yourself. Run it on a home server for a private 'company chatbot' without per-seat SaaS pricing.

The catch

Jump to setup guide ↓

You maintain updates and security; initial Docker setup takes effort.

How to set up Open WebUI

Browser chat UI on your own machine, wired to Ollama or API keys—private team/family AI without per-seat SaaS.

Time: 30–45 min
Difficulty: Technical
Verified: 2026-06-22

Before you start

Docker Desktop installed (Mac/Windows) or Docker on Linux
Ollama running locally if you want free local models (see our Ollama setup guide)
Port 3000 free on the host

Pull and run with Docker
One-liner from official docs: docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main. On Linux without host.docker.internal, use --network host or point Ollama URL to your LAN IP.
Create the admin account
Open http://localhost:3000. First user to sign up becomes admin—use a strong password. This instance is yours; no cloud account required.
Connect Ollama
Settings → Connections → Ollama. Base URL: http://host.docker.internal:11434 (Docker Desktop) or http://127.0.0.1:11434 (Linux host network). Save, then refresh model list—you should see models you pulled in Ollama.
Send a test message
New chat → pick a model (e.g. llama3.2:3b) → ask something real. If models are empty, Ollama isn't reachable from the container—fix the URL before inviting anyone else.
Optional — add BYOK cloud models
Settings → Connections → OpenAI (or compatible). Paste your API key and base URL if using a proxy. Use for harder tasks; keep Ollama as the free default for daily chat.
Optional — invite household users
Admin Panel → Users. Create accounts for family or teammates. Everyone hits your server—not a rented multi-seat plan.

Troubleshooting

No models appear: Ollama not running or wrong URL from inside Docker. Test curl http://host.docker.internal:11434/api/tags from the container shell.
Container won't start: Port 3000 in use—change -p 3001:8080 and open :3001 instead. Check Docker has enough disk for the image.
Slow responses: Model too large for host RAM or CPU. Use a smaller Ollama tag or move Open WebUI to a machine with more headroom.

Keep it working

docker pull ghcr.io/open-webui/open-webui:main monthly, then recreate container to patch
Back up the open-webui Docker volume—it holds users and chat history
Don't expose port 3000 to the public internet without auth and HTTPS

Official docs: docs.openwebui.com/getting-started/quick-start/