Self-HostingIntermediate25 minVerified 45 days ago

Deploy Open WebUI with Ollama

Run a self-hosted ChatGPT-style interface that talks to your local Ollama models and any OpenAI-compatible API.

open-webuiollamaself-hostingchatdocker

The promise

Every powerful local model deserves a good interface. Open WebUI gives you a ChatGPT-style web app that runs on your own hardware, connects to Ollama, and can even bridge to cloud APIs for side-by-side comparison. It is the fastest way to turn a raw Ollama server into something the rest of your team will actually use.

This recipe deploys Open WebUI with Docker, connects it to a local Ollama instance, and shows you how to add a cloud provider so you can compare local and frontier models in the same conversation.

What you'll get

Open WebUI running in Docker on port 3000
Connection to your local Ollama instance
One or more local models available in the dropdown
Optional cloud API bridge for comparison

Prerequisites

Ollama running locally (see the Ollama CUDA recipe)
Docker installed
At least 2 GB free disk space for the Open WebUI image
One pulled Ollama model, such as qwen3.5:9b

Sanity checks

Check	Command / Action
Container running	`docker ps`
Ollama reachable	`curl http://localhost:11434/api/tags`
WebUI logs	`docker logs open-webui`
Model list loads	Refresh the browser and check the dropdown

Common gotchas

Symptom	Fix
"Cannot connect to Ollama"	Verify `--add-host=host.docker.internal:host-gateway` and that the Ollama endpoint is `http://host.docker.internal:11434`.
No models listed	Pull at least one model with `ollama pull qwen3.5:9b`.
Slow first load	The Docker image is large. Allow several minutes for the initial pull.
Admin lockout	The first user to sign up is admin. There is no password recovery without database access.

Next step

Use Open WebUI as a playground to test prompts before using them with agents like Cline, Aider, or OpenClaw. Then move the best prompts into versioned prompt files for your agents.

Steps

One command starts the container and persists its data:

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

Visit http://localhost:3000. The first user to sign up becomes the admin.

The first time you open Open WebUI, it asks for an Ollama API URL:

http://host.docker.internal:11434

Use that exact value if Ollama is running on the Docker host. If Ollama is also in Docker, put both containers on the same network and use the Ollama container name as the host.

Click Save. The models you have pulled appear in the model selector.

Open WebUI can pull Ollama models directly without touching the terminal:

Go to Settings → Models.
Click Pull a model.
Enter qwen3.5:9b and wait.

You can also pull from the admin panel in bulk.

This is where Open WebUI becomes more than a local chat app:

Go to Settings → Connections.
Add an OpenAI-compatible API base URL and key:
- OpenAI: https://api.openai.com/v1
- Anthropic: use the Anthropic connector
- OpenRouter: https://openrouter.ai/api/v1
Save. The new provider appears in the model dropdown.

Now you can ask the same question to a local model and a cloud model and compare the answers.

To let other devices on your network reach Open WebUI:

docker run -d -p 3000:8080 \
  -e ENV=prod \
  -e OLLAMA_BASE_URL=http://host.docker.internal:11434 \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Then access it at http://your-lan-ip:3000.

Do not expose Open WebUI to the public internet without authentication. The first signup becomes admin, and there is no rate limit by default.

Recipe verified Mon Jun 15 2026 00:00:00 GMT+0000 (Coordinated Universal Time). Commands are tested but your environment may differ.

Browse related services