Key Takeaways
- Open WebUI (openwebui.com) is a free, self-hosted web app with over 150,000 GitHub stars.
- It started in October 2023 as "Ollama WebUI" and was renamed to "Open WebUI" in early 2024 after trademark concerns raised by the Ollama team.
- It connects to Ollama, vLLM, llama.cpp, LiteLLM, and any OpenAI-compatible API β it does not run models itself.
- Built-in features include document RAG, multi-user authentication with roles, a pipelines/functions plugin system, and image, voice, and web-search integrations.
- The license changed from BSD-3-Clause to a custom "Open WebUI License" in April 2025, which adds a branding clause above 50 users unless you buy an Enterprise license.
π In One Sentence
Open WebUI is a free, self-hosted, multi-user web interface that connects to Ollama, vLLM, llama.cpp, or any OpenAI-compatible API, adding built-in RAG, plugins, and role-based access on top.
π¬ In Plain Terms
Open WebUI doesn't run AI models itself β think of it as the ChatGPT website you install on your own server, in front of models you're already running or paying for elsewhere.
Quick Facts: Open WebUI
- License: "Open WebUI License" β BSD-derived, source-available, with a branding clause above 50 users/30 days (not OSI-approved open source)
- Platforms: Self-hosted web app (Docker, pip, or Kubernetes); works in any browser; PWA install on mobile
- Website: openwebui.com
- GitHub: github.com/open-webui/open-webui β 150,000+ stars, 22,000+ forks
- Cost: Free to self-host; paid Enterprise license available for white-labeling, SSO, and support
- Core function: Self-hosted, multi-user chat interface for Ollama and OpenAI-compatible APIs, with built-in RAG
What Is Open WebUI?
Open WebUI is a free, self-hosted web application that gives you a ChatGPT-style chat interface for models you run yourself or connect to via API. You install it on your own machine, home server, or company server with a single Docker command, and it serves a web page that anyone on your network β or the internet, if you choose to expose it β can log into and chat through.
It does not train or run models on its own. Instead, it is a front end that talks to a backend: most commonly Ollama, but also vLLM, llama.cpp's server mode, LiteLLM, or any endpoint that speaks the OpenAI or Anthropic API format, including cloud providers if you add an API key.
What separates it from single-user desktop clients is that it was built as a shared, multi-user system from early on: accounts, roles, and permissions are core features, not an afterthought, which is why it is popular for households, small teams, and homelab setups where more than one person needs access to the same models.
History: From "Ollama WebUI" to Open WebUI
Open WebUI was created by developer Timothy Jaeryang Baek ("tjbck" on GitHub), with the first commit to the repository landing on October 6, 2023. The project was originally named "Ollama WebUI" β a simple, honest name for what it was: a web interface built specifically to sit in front of Ollama.
As the project grew popular and started adding support for backends beyond Ollama, the Ollama team itself raised concerns that the name and a related "Ollama Hub" branding were creating public confusion about which projects were official Ollama products. Rather than a legal dispute, this played out as a public conversation: the maintainers opened a GitHub discussion, the community voted between candidate names, and the project was renamed "Open WebUI" around January 2024 β a name that also better reflected its broader, backend-agnostic direction.
Since the rename, the project has shipped rapidly: version 0.9.x added knowledge-base folders and SSRF hardening; 0.10.x added folder sharing, automatic context compaction, and a memory overhaul; and by version 0.11.x (mid-to-late 2026) it had a redesigned UI plus "sub-agents" that can delegate background tasks and request human approval before using a tool. As of early September 2026, the current stable release is v0.11.3, published August 31, 2026.
A more consequential change than any single feature was the license switch in version 0.6.6, released April 19, 2025: code was BSD-3-Clause through v0.6.5, then moved to a custom "Open WebUI License" β still permissive and free to self-host, but with a clause requiring visible "Open WebUI" branding in any deployment, unless it serves under 50 users in a rolling 30-day window, has written permission, or is covered by a paid Enterprise license.
What Can You Actually Do With Open WebUI?
- Chat with Ollama or any OpenAI-compatible model. Point it at a local Ollama install, a vLLM or llama.cpp server, or a cloud API key, and every connected model shows up in the same model picker.
- Chat with your own documents (RAG). Upload PDFs, text, or other files into a knowledge collection and Open WebUI retrieves relevant passages before answering, with a choice of vector database backends.
- Run it for a household or a team, not just yourself. Built-in multi-user authentication with roles (admin, user, pending) means everyone gets their own login and chat history on one shared server.
- Extend it with pipelines and functions. A plugin framework called Pipelines lets you inject custom logic β content filters, external tool calls, or custom RAG pipelines β into the request/response flow.
- Generate images, use voice, and search the web from chat. Open WebUI integrates image generation backends, voice and video calling, and web search providers directly into the conversation.
- Install it as an app on your phone. As a Progressive Web App, it can be added to a mobile home screen and used like a native app without going through an app store.
Who Should Use Open WebUI?
Open WebUI is built for people who want a shared, private ChatGPT-style front end running on hardware they control β not for someone who wants a single-user desktop app with nothing to configure.
- β You run Ollama and want a browser-based UI instead of the terminal β Open WebUI turns `ollama run` into a full chat app with history, uploads, and settings.
- β More than one person needs to use the same models β built-in accounts and roles mean you are not sharing one login or one machine.
- β You want to chat with your own documents without a separate RAG tool β the built-in knowledge feature covers this out of the box.
- β You are comfortable running Docker or a small server β the standard install is one Docker command, but it does assume basic self-hosting comfort.
- β You want zero server setup and a single-user app β a desktop client like LM Studio or Cherry Studio needs no server at all.
- If unsure: install Ollama first, then run Open WebUI's single Docker command against it β you can be chatting through a browser in under ten minutes.
When Should You NOT Use Open WebUI?
Being a server-based, multi-user tool is Open WebUI's biggest strength and also the source of its main limitations.
- If you want a single-click desktop app with no server to maintain: Open WebUI is a web service you run and keep updated yourself; a desktop client avoids that entirely.
- If you need to white-label the interface for external customers at scale: the "Open WebUI License" branding clause applies above 50 users in 30 days unless you purchase an Enterprise license.
- If you expect it to run models on its own: it has no built-in inference engine β you must separately install and maintain Ollama, vLLM, or another backend.
- If you plan to expose it to the public internet without hardening it: like any self-hosted web app, it needs a reverse proxy, TLS, and up-to-date patches; check the changelog for security fixes before deploying to production.
- If you want guaranteed commercial support with an SLA: that requires the paid Enterprise license β the free tier is community-supported only.
How Do You Install Open WebUI?
Setup has one prerequisite and two supported install paths:
- 1. Install a backend first. Open WebUI needs something to talk to β most people install Ollama first (`ollama pull llama3.2` to grab a model), but any OpenAI-compatible endpoint works too.
- 2. Recommended: run it with Docker. `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` β add `--gpus all` and swap in the `:cuda` image tag if you have an NVIDIA GPU.
- 3. Alternative: install with pip. `pip install open-webui` then `open-webui serve`, using Python 3.11 or 3.12 (3.13 is not yet supported as of this writing).
- 4. Create the first account. Open `http://localhost:3000` and sign up β the first account created automatically becomes the admin.
- 5. Connect models and invite others. Add connections under Settings β Connections, then invite users from the Admin panel and assign roles.
- Full documentation, Docker Compose files, and Kubernetes/Helm charts are maintained at docs.openwebui.com.
Pricing: Free Core, Paid Enterprise License
Open WebUI follows an open-core-adjacent model: the software is free to self-host, and a separate paid tier covers commercial and organizational needs.
- Free (self-hosted): The full application β chat, RAG, multi-user auth, pipelines, image/voice/web-search integrations β is free under the source-available "Open WebUI License." You provide your own server and backend (Ollama, vLLM, or a cloud API key you pay for separately).
- Enterprise license (custom pricing): Aimed at organizations, not individuals. Covers white-labeling and branding removal above the 50-user threshold, SSO/LDAP/Active Directory integration, high-availability multi-node deployment, compliance support referencing SOC 2, HIPAA, and GDPR, and dedicated support with an SLA. Pricing is not publicly listed β it is quote-based, contact-sales only, so treat any third-party "price estimate" you see elsewhere with caution until you get a quote directly.
- No official hosted/managed SaaS version. As of this writing, the maintainers position self-hosting as the primary model rather than offering their own cloud-hosted instance you sign up for directly.
- Optional support: the project also accepts sponsorships through GitHub Sponsors for those who want to fund development without buying an Enterprise seat.
Open WebUI vs LibreChat vs LobeChat vs AnythingLLM: Which One?
All four are free, self-hostable chat front ends, but each was built around a different center of gravity.
Tool | Link | Best for | Runs models itself | Built-in RAG | Multi-user auth | License |
|---|---|---|---|---|---|---|
| Open WebUI | openwebui.com | Self-hosted, multi-user front end for Ollama and OpenAI-compatible APIs | No (needs Ollama/vLLM/API) | Yes | Yes, with roles | Open WebUI License (BSD-derived) |
| LibreChat | librechat.ai | Managing many cloud provider API keys behind one shared UI | No | Yes | Yes, with roles | MIT |
| LobeChat | lobehub.com | Polished consumer-style UI with a plugin marketplace | No | Yes | Yes | Apache 2.0 |
| AnythingLLM | anythingllm.com | RAG-first document chat with a vector store as the core concept | No (via Ollama/LM Studio) | Yes (core feature) | Yes | MIT |
For single-user desktop clients rather than self-hosted servers, see Cherry Studio and Best Local LLM Frontends. For character/roleplay-focused chat instead, see Open WebUI vs SillyTavern.
Common Mistakes With Open WebUI
- Installing Open WebUI before installing a backend. It has nothing to talk to until Ollama, vLLM, or another compatible endpoint is running and reachable.
- Exposing it to the internet without a reverse proxy or TLS. Treat it like any other self-hosted web app β put it behind Nginx/Caddy/Traefik with HTTPS before making it reachable outside your own network.
- Assuming the free tier includes white-label branding removal at any scale. Above 50 users in a rolling 30-day period, the license requires either keeping "Open WebUI" branding visible or purchasing an Enterprise license.
- Confusing it with Ollama itself. Open WebUI is the chat interface; Ollama (or vLLM, llama.cpp) is the separate model-running engine underneath it.
- Skipping updates. Because it is exposed as a network service, staying on old versions carries more risk than with a desktop app β check the changelog and update regularly.
Common Questions About Open WebUI
What is Open WebUI?
Open WebUI is a free, self-hosted web interface for chatting with Ollama, vLLM, llama.cpp, or any OpenAI-compatible API. It adds a multi-user ChatGPT-style front end, document RAG, and a plugin system on top of a model backend you run or connect yourself.
Is Open WebUI free?
Yes, the core project is free to self-host. Since version 0.6.6 (April 2025) it ships under the custom "Open WebUI License" rather than plain BSD-3-Clause β a BSD-style license with a branding clause. A separate paid Enterprise license exists for organizations that want white-labeling, SSO, or dedicated support.
What was Open WebUI originally called?
It launched in October 2023 as "Ollama WebUI." It was renamed to "Open WebUI" around January 2024 after the Ollama team raised trademark-confusion concerns, and the community voted on the new name in a public GitHub discussion.
Does Open WebUI require Ollama?
No. Ollama is the most common backend, but Open WebUI also connects to vLLM, llama.cpp's server mode, LiteLLM, and any OpenAI- or Anthropic-compatible API, including cloud providers.
Can multiple people use the same Open WebUI instance?
Yes. Open WebUI has built-in multi-user authentication with role-based access control (admin, user, pending), which is one of its main differences from single-user desktop clients.
Does Open WebUI have RAG (chat with documents)?
Yes. Open WebUI includes built-in retrieval-augmented generation with a choice of vector database backends, letting you upload documents and ask questions grounded in their content.
Is the "Open WebUI License" still open source?
It is source-available and free to self-host, but it is not an OSI-approved open-source license. The license requires keeping visible "Open WebUI" branding in any distributed or hosted deployment above 50 users in a 30-day period, unless you have written permission or an Enterprise license.
Does Open WebUI run models itself?
No. Open WebUI is an interface layer, not an inference engine. It needs a backend β Ollama, vLLM, llama.cpp, or an API-compatible endpoint β to actually generate responses.
How is Open WebUI different from LibreChat or LobeChat?
All three are self-hosted, open-source chat front ends, but Open WebUI is the most tightly integrated with Ollama and local-first workflows, while LibreChat leans toward multi-provider cloud API management and LobeChat emphasizes a polished consumer-style UI and plugin marketplace.
Is Open WebUI safe to expose to the internet?
It can be, but like any self-hosted web app it needs a reverse proxy, TLS, and authentication hardening. Check the current changelog for security-relevant fixes and keep the container updated before exposing an instance publicly.
Sources
- Open WebUI official site and documentation β openwebui.com, docs.openwebui.com
- Open WebUI source code, releases, and license β github.com/open-webui/open-webui
- GitHub Discussion #602 β community rename vote from "Ollama WebUI" to "Open WebUI"
- Open WebUI Enterprise licensing page β docs.openwebui.com/enterprise