Blog · Tutorials · September 18, 2026 · 7 min read

Build Your Own Private ChatGPT: Ollama + Open WebUI on Any PC

Self-hosted chat interface running on local hardware

Ollama’s terminal chat is great for testing, but nobody wants to live in a command line. Open WebUI is a free, self-hosted interface that turns Ollama into something that looks and feels like ChatGPT — conversation history, model switching, document uploads, and even separate accounts for family or teammates. All of it on your hardware, with nothing leaving your machine.

Prerequisites

Ollama installed and working (see our getting-started guide).
Docker Desktop (Windows/macOS) or Docker Engine (Linux) — the simplest way to run Open WebUI. A non-Docker pip install also exists if you prefer Python.
About 2 GB of free disk for the WebUI container itself.

Step 1: Start Open WebUI

With Ollama already running, start the WebUI and point it at Ollama’s API:

docker run -d -p 3000:8080 \
  -v open-webui:/app/backend/data \
  --add-host=host.docker.internal:host-gateway \
  --name open-webui --restart always \
  ghcr.io/open-webui/open-webui:main

On the same machine, Docker’s host.docker.internal address reaches Ollama at port 11434 automatically. Open http://localhost:3000 in your browser.

Step 2: Create your account

The first account registered on a fresh install becomes the administrator. Because everything is stored in your own Docker volume, the credentials never leave your machine — this is your private instance, not a cloud service.

Step 3: Chat like ChatGPT

Pick a model from the dropdown — every Ollama model you have pulled appears automatically.
Full history: conversations are saved, searchable, and can be renamed or organized into folders.
Document chat: attach a PDF or text file and the UI indexes it for retrieval — ask questions about your own documents entirely offline.
System prompts: save custom personas (code reviewer, translator, writing coach) as reusable presets.

Step 4 (optional): Share it on your home network

Other devices on your Wi-Fi can reach the UI at http://<your-pc-ip>:3000 — useful for giving a laptop or tablet access to the same local models. Keep it inside your network: exposing this port to the internet without setting up authentication and HTTPS properly is a genuine security risk, so don’t port-forward it casually.

Everyday tips

In WebUI’s model settings you can raise the context length per conversation — before cranking it up, check what your VRAM can take in our compatibility panel, since context costs memory linearly.
Downloaded but rarely used models still occupy disk; prune them with ollama rm from time to time.
Curious whether all this local usage would actually cost less than a subscription or API metering? The Local vs Cloud panel gives you the honest math including electricity.

That is the whole setup: one Docker command on top of Ollama, and you have a private, capable assistant with no metering, no logs on someone else’s server, and no monthly bill.