Menu
Open source

Open WebUI

4 min readbeginnerUpdated 28 Sept 2026
1 · In one line

Open WebUI, a free chat app you host yourself, lets you talk to AI models running on your own computer or in the cloud, from one web page.

1 · What it is

Open WebUI is a chat app you run yourself. It can live on your own computer or on a server, a computer that stays switched on for others to reach. You open it in a web browser and chat there, and you decide which AI models sit behind it. It can even be set up to work with no internet connection at all.

Open WebUI does not do the thinking itself. It connects to model providers: programs or services that actually run a model. One kind is a local server such as Ollama, which runs downloaded models on your own machine. Another is a cloud service. You reach it through an API, a door that lets programs talk to each other. You add a provider by typing its address and a key, and its models appear in a drop-down list.

Picture asking a question with a small model on your laptop, then trying the same question on a large cloud model to compare. Both sit in the same list. There is a catch. The model you pick decides where your message travels. A cloud model sends your words to that company. You can also upload your own files into a “knowledge base”, and the app searches them for passages that match your question. The code is free to use and change. But since version 0.6.6, the licence has one rule for setups with more than 50 users: they must keep the Open WebUI branding.

2 · Why it exists

Running your own models is one job, and chatting with them comfortably is another.

No friendly screenA local model server runs the model, but you still want a chat window to talk to it.
Too many appsLocal models and cloud models usually live in separate apps, and Open WebUI puts them in one list.
Your own filesYou often want answers from your own documents, and Open WebUI can search uploaded files for you.
3 · How it works

Follow one message from your keyboard to a reply.

How Open WebUI routes one chat message You type a message in the browser. The highlighted step, Open WebUI, sends it to the address of the model you picked. That is either a local server such as Ollama on your machine or a cloud API. The provider runs the model and the reply streams back into the chat page. ONE MESSAGE, ONE CHOSEN MODEL, ONE REPLY Your message Open WebUI Local server Cloud API Reply typed in the chat page in your browser sends it to the picked model's address e.g. Ollama, on your own machine a company's servers run the model streams back into the chat, word by word Only one path is used: a local model runs on your machine, a cloud model sends your words away.
The highlighted box is Open WebUI: it chooses where your message goes but does not run the model itself.
  1. 1 · typeYou pick a model from the list and type a message in the chat page.
  2. 2 · routeOpen WebUI sends the message to the address of that model's provider.
  3. 3 · runThe provider, local or in the cloud, runs the model on your message.
  4. 4 · streamThe reply flows back word by word and appears in the chat.

The model you pick decides where your words are sent, so a cloud model sees the message and a local one runs the reply on your machine.

4 · Where it's used
WhoWhat they askWhat it works with
Student“Can I chat with an open model on my laptop in a normal browser window?”A local Ollama model shown in the Open WebUI model list
Small team“Can we share one AI chat page but control who sees which models?”Per-group access rules on model presets
Researcher“Which of these PDFs mentions the budget, and where?”A knowledge base of uploaded files
Tinkerer“Does a local model answer this as well as a cloud one?”The same prompt sent to two connected models
5 · What it solves, and what it doesn't
solves
  • It gives local and cloud models one shared chat page.
  • It connects to any server that speaks the Ollama or OpenAI-compatible protocol.
  • It can search your uploaded documents and cite where it found an answer.
  • It lets an administrator limit which users or groups see each model.
doesn't solve
  • It does not run models itself; a separate provider does that work.
  • Picking a cloud model sends your prompt to that cloud company.
  • Using a local model does not make separately set-up cloud services local.
  • Its licence requires deployments with more than 50 users to keep the Open WebUI branding.
6 · Go deeper
Ollamallama.cppLM StudioRAG

Sources used

This explainer is written in original language. The links below support its factual claims.

  1. repoopen-webui/open-webui: User-friendly AI Interface, Open WebUI · read 28 Sept 2026
  2. docsOpen WebUI documentation, Open WebUI · read 28 Sept 2026
  3. docsConnect Local and Cloud Models, Open WebUI · read 28 Sept 2026
  4. docsKnowledge Bases and Document Chat, Open WebUI · read 28 Sept 2026
  5. docsModels, Open WebUI · read 28 Sept 2026
  6. officialOpen WebUI License, Open WebUI · read 28 Sept 2026
Suggest a correction