How to Set Up Open WebUI with Ollama in 2026

Quick answer

Open WebUI gives your local Ollama models a polished, ChatGPT-style web interface that runs entirely on your machine. With Docker installed and Ollama running, start it with one command: 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, then open http://localhost:3000. It adds conversation history, multi-model switching, and document chat (RAG) on top of Ollama — all offline.

Ollama’s command line is great for running models, but most people want a real chat interface. Open WebUI is the most popular self-hosted option: it looks and feels like ChatGPT, connects to your local Ollama models, and keeps everything on your hardware. This guide gets it running and connected.

If you have not installed Ollama yet, do that first with our Ollama install guide.

What You Need

Ollama installed and running, with at least one model pulled (e.g. ollama run llama3.1:8b once, then exit).
Docker installed — Open WebUI runs in a container, which is the cleanest way to install it.
A few GB of free disk space for the container and its data.

Step-By-Step Setup

1. Confirm Ollama is running

Open WebUI talks to Ollama’s API (port 11434). Make sure Ollama is installed and you have pulled a model:

ollama list

If that shows at least one model, you are ready.

2. Start Open WebUI with Docker

Run the official container. This version connects to the Ollama already running on your machine:

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

What the flags do: -p 3000:8080 exposes the UI on port 3000, --add-host lets the container reach Ollama on your host, -v open-webui:... persists your chats and settings, and --restart always brings it back after a reboot.

Prefer an all-in-one install? There is a bundled image that includes Ollama itself (ghcr.io/open-webui/open-webui:ollama) — handy if you want everything in one container, though running Ollama natively is usually smoother for GPU access.

3. Open the interface

Go to http://localhost:3000 in your browser. On first launch you create a local admin account — this stays on your machine; it is not a cloud login.

4. Select a model and chat

Your Ollama models appear in the model picker at the top. Choose one (e.g. llama3.1:8b), type a message, and you are chatting with a local model through a clean interface — conversation history, markdown, code blocks, and all.

The Features Worth Knowing

Open WebUI does more than chat:

Document chat (RAG): Click the paperclip and upload a PDF, Word doc, or text file. Open WebUI chunks and indexes it, then your local model answers questions about its contents — all offline.
Multi-model switching: Jump between models mid-conversation to compare answers.
Conversation history: Every chat is saved and searchable.
Web search, image generation, voice, and tools: Optional integrations and Pipelines/Functions extend it well beyond a basic chat box.
Multi-user: Host it for a household or team, each with their own account.

These features — especially native RAG — are why Open WebUI is the default upgrade from the Ollama CLI. For the CLI-vs-GUI tradeoff with the other popular tool, see LM Studio vs Ollama.

Common Mistakes

Open WebUI can’t see your models. This almost always means the container can’t reach Ollama. Confirm Ollama is running and that you included the --add-host=host.docker.internal:host-gateway flag.
Forgetting to pull a model first. Open WebUI is an interface, not a model source for Ollama — pull at least one model with ollama pull (or run) before expecting it in the picker.
Expecting a cloud account. The first-run account is local only. Nothing leaves your machine.
GPU not used inside the all-in-one image. If you use the bundled :ollama image, GPU passthrough needs extra Docker flags — running Ollama natively and Open WebUI in its own container is the simpler path.

What To Do Next

Pick good models to run through it: best local LLM models 2026.
Try RAG with your own documents (the paperclip upload).
Make sure your hardware fits the models you want: VRAM requirements guide.
Compare with the desktop-app alternative: LM Studio vs Ollama.

Frequently Asked Questions

What is Open WebUI?

Open WebUI is a free, open-source, self-hosted web interface for local LLMs. It looks and feels like ChatGPT but connects to your local Ollama models and runs entirely on your hardware, so your prompts and chats never leave your machine.

How do I connect Open WebUI to Ollama?

Run Open WebUI in Docker with the --add-host=host.docker.internal:host-gateway flag so the container can reach Ollama’s API (port 11434) on your host. Once both are running, your Ollama models appear automatically in Open WebUI’s model picker.

Does Open WebUI support RAG / document chat?

Yes, out of the box. Click the paperclip in the chat input to upload a PDF, Word, or text file; Open WebUI chunks and indexes it, then your local model answers questions about the content — all offline.

Is Open WebUI free and private?

Yes. It is free and open source, and everything runs locally — prompts, responses, conversation history, and uploaded documents all stay on your machine. The first-run account is local, not a cloud login.

Open WebUI or LM Studio?

Open WebUI is a self-hosted web interface that pairs with Ollama and shines for RAG, multi-user hosting, and tooling. LM Studio is a desktop app with built-in model browsing and fast Apple Silicon inference. See LM Studio vs Ollama; many people use one of each.

Last updated: June 2026. Setup commands reference the official Open WebUI Docker images. Features reflect the 2026 release. We do not publish invented benchmarks.

How to Set Up Open WebUI — A Private ChatGPT for Local Models (2026)

Prerequisites