Open WebUI and Ollama on Your Own PC: Do You Need Two Services
How to install Open WebUI through Docker or pip, how to connect Ollama through OLLAMA_BASE_URL, and at what point a web interface on top of a local model stops paying for itself.
Ollama runs in the terminal with no interface. Open WebUI gives you a browser chat that looks like ChatGPT. The pairing makes sense on paper, and almost every guide you'll find describes exactly this combination.
Almost all of those guides, though, are written by hosting companies and end with a pitch to run everything in the cloud. For a home machine, the math works out differently, and it's worth saying so before you install anything.
How to install it
Two ways, both official.
Through Docker:
docker run -d -p 3000:8080 -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:main
Through Python, if you'd rather not install Docker:
pip install open-webui
open-webui serve
The interface opens at http://localhost:3000. Inside the container, the app
listens on 8080 — hence the odd-looking pair of ports in the command.
There's also a build with Open WebUI and Ollama bundled in one container, tagged
:ollama. It saves you the trouble of linking two services, but ties updating one
to the other.
How to connect Ollama
If Ollama runs on the same machine, Open WebUI usually finds it on its own. If not, the address is set through an environment variable:
docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=http://192.168.1.10:11434 -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:main
The most common mistake at this step: putting localhost in the container's
variable. Inside a container, localhost means the container itself, not your
computer. You need either the machine's address on the network, or the special
hostname Docker exposes for it.
How Ollama differs from LM Studio as a model source is covered in a separate breakdown.
What you get
The interface is genuinely good, and it's not just a chat window. There's a knowledge base with document upload and search, notes, channels, model management, access control, and an admin panel. Plus extensibility: Python tools and functions, markdown-based skills, pipelines, and MCP support. There's a separate article on that part.
The feature set is impressive. The problem is it's built for a different scenario.
Who this is actually for
Open WebUI is designed as a multi-user platform. Authentication, roles, model-access permissions, scaling, production observability — all of this makes sense when a team sits behind one server.
If there's just one user, half the capabilities turn into extra screens. You're administering a platform in order to talk to a model that's already installed on that same computer.
There's also an architectural cost. You now have two services instead of one: Ollama holds the model, Open WebUI holds the interface and its own database. Both need updating, both can break independently, and a typical failure looks like "models aren't showing up in the list" — with the actual cause somewhere in the network between containers.
When the pairing is worth it
It's worth it in three cases. First: several people use the model, and you need accounts and permissions. Second: the GPU server sits separately, and you connect from a laptop. Third: you're deliberately building a platform and need pipelines and your own Python logic.
In all three, Open WebUI is a strong, arguably the best, open option.
When it isn't worth it
If there's one user and one machine, you're installing two programs to get an interface you could get with one. And what you get is still a chat — meaning you're still copying results into files by hand.
Doka covers the same scenario with a single app: a local model installs from the interface, nothing to link together, and the work happens not in a chat window but directly with files and the terminal. MCP servers connect in the same place, and tasks can be put on a schedule.
The honest test is this: if you need a web interface because different people connect to the model from different devices, install Open WebUI without a second thought. If you only need a web interface because Ollama doesn't have one of its own, that's an unnecessary layer, and a desktop app solves it in fewer steps.