How to Use Open WebUI

Open WebUI gives you a ChatGPT-style interface on top of your own local models, with RAG and MCP built in.

Open WebUI (formerly Ollama WebUI) is the most polished self-hosted chat front-end in the open-source ecosystem. It targets Ollama by default but speaks any OpenAI-compatible API, bundles document RAG, and recently added MCP tool support. One Docker command gets you a private ChatGPT clone that runs entirely on your hardware.

What it is

A SvelteKit web app backed by a FastAPI service. It manages users, chat history, model catalogs, prompt libraries, and a built-in vector store for retrieval. It connects to Ollama at http://host.docker.internal:11434 out of the box, and to any OpenAI-compatible endpoint through a config panel.

Install / set up

# docker deploy
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

First run

Open http://localhost:3000, create the admin account (the first signup is always admin), and pick a model from the dropdown. If Ollama is running locally with at least one pulled model, it appears automatically.

$ ollama pull llama3.1
$ curl http://localhost:3000
# then chat in the browser

Everyday workflows

• Drop a PDF into the chat to index it on the fly and ask questions against it — RAG happens inline.
• Create a “model” preset with a system prompt, temperature, and knowledge collection, then share it with your team.
• Register an MCP server in Settings → Tools to let the chat call external APIs (GitHub, filesystem, databases).

Gotchas and tips

The first signup becomes admin, so register yourself before exposing the port. If you skip this step and a stranger hits the URL first, they own the instance. For production, set WEBUI_AUTH=true and put it behind a reverse proxy with TLS.

Updates ship fast — sometimes multiple times a week. Pin a specific image tag in production rather than chasing main, and read the release notes before upgrading because database migrations are common.

Who it’s for

Teams and hobbyists who want the ChatGPT UX on local or private models. If you run Ollama and you’re tired of the terminal, this is the front-end you install next.