Open WebUI

What does Open WebUI do?

Open WebUI describes itself as "a home for AI, a self-hosted AI platform that's extensible, feature-rich, user-friendly, and built to run entirely offline." It connects to "Ollama and OpenAI-compatible APIs" for local and cloud models, with LM Studio, GroqCloud, Mistral, OpenRouter and vLLM named as endpoints. Document search uses "Retrieval Augmented Generation backed by 9 vector databases", including ChromaDB, PGVector, Qdrant, Milvus and Pinecone, with hybrid keyword and vector search and re-ranking. Web search runs through dozens of providers including SearXNG, Brave, Kagi, Tavily and DuckDuckGo, and images come from DALL-E, Gemini, or local ComfyUI and AUTOMATIC1111.

It is built for teams. Administrators define roles, groups and permissions; sign-in supports LDAP and Active Directory, OAuth single sign-on, and SCIM 2.0 provisioning from Okta, Azure AD and Google Workspace. Plugins come as filters, actions, pipes, tools and skills, and external tools connect through MCP, MCPO and OpenAPI servers. Voice and video calls use local Whisper or hosted speech services. It also lists shared channels, memory across chats, scheduled prompts, side-by-side model chats, an arena with rankings, and admin dashboards for token use and cost. Storage is SQLite or PostgreSQL, with files on disk or in S3, Google Cloud or Azure.

Key facts

• Licence: the Open WebUI License, BSD-3-Clause text plus a clause 4 that forbids altering or removing the Open WebUI branding, except for deployments of 50 or fewer users in any rolling 30 days, with written permission, or under a paid enterprise licence.
• Contributions fall under a Contributor License Agreement; older material keeps its earlier licence, per LICENSE_HISTORY. GitHub reports the licence as Other.
• Backends: Ollama and any OpenAI-compatible API; the README names LM Studio, GroqCloud, Mistral, OpenRouter and vLLM.
• Document search over 9 vector databases with hybrid search and re-ranking; web search through dozens of providers including SearXNG.
• Teams: roles, groups and permissions; LDAP and Active Directory; OAuth single sign-on; SCIM 2.0 provisioning.
• Tools through MCP, MCPO and OpenAPI servers; plugins as filters, actions, pipes, tools and skills.
• Install: pip install open-webui with Python 3.11, then open-webui serve on port 8080; Docker images main, cuda and ollama (Ollama bundled); Kubernetes through kubectl, kustomize or Helm.
• Latest tagged release when read: v0.11.3 on 2026-08-31; 152,423 stars on 2026-09-17, the most of any chat interface in this directory.

What does it replace, and where does it fall short?

Open WebUI replaces ChatGPT and Claude for a team that wants one chat interface over its own models and any hosted ones, with its documents, web search, roles and single sign-on, on its own servers. It is the most complete team platform on the open-source alternatives to ChatGPT page.

Where it falls short: the licence is source available. Above 50 users you may not remove or change the Open WebUI branding without written permission or a paid enterprise licence, and the README says doing so is a material breach. The README does not name Anthropic as a backend; Claude models need an OpenAI-compatible gateway in front of them. For a plain MIT platform with Anthropic built in, LibreChat.

How does Open WebUI run?

pip install open-webui on Python 3.11, then open-webui serve, which listens on port 8080. Or run a Docker image: main, cuda for NVIDIA GPUs, or ollama with Ollama bundled inside. Kubernetes installs through kubectl, kustomize or Helm. Mount the data volume, or the README warns of data loss. Offline mode is one environment variable. The maintainers sell an enterprise plan through a sales team; the README describes no hosted cloud.

Who is Open WebUI for?

A company that wants a private ChatGPT for its staff, with its own documents, sign-on and roles, and is either under 50 users or content to keep the Open WebUI name on it. A team that must rebrand it without paying should pick LibreChat or AnythingLLM, both MIT.

What limits does the README state?

From the README: some Docker environments need extra configuration; use Python 3.11 to avoid compatibility issues; the dev branch "may have bugs or incomplete features"; leaving out the data volume risks losing data. The licence's branding clause, quoted above, is the limit to read first.

Questions people ask

Is Open WebUI open source?

Not by the OSI definition. Its licence is BSD-3-Clause plus a clause that forbids altering or removing the Open WebUI branding, except for deployments of 50 or fewer users in any rolling 30-day window, with written permission, or under a paid enterprise licence. A restriction on how you may modify the software is why this site labels it source available.

Can I use Open WebUI for my company?

Yes. The licence permits use and modification; the branding clause is the condition. Keep the Open WebUI name and logo and there is no user limit. Rebrand it and you need 50 or fewer users in any 30 days, written permission, or an enterprise licence from Open WebUI Inc.

Does Open WebUI work with Claude or Anthropic models?

The README names Ollama and OpenAI-compatible APIs, with LM Studio, GroqCloud, Mistral, OpenRouter and vLLM as examples, and does not name Anthropic. Claude models therefore need an OpenAI-compatible gateway such as OpenRouter in front of them. LibreChat lists Anthropic directly if that matters.

How does Open WebUI compare with ChatGPT?

It gives a team ChatGPT's shape, chat, documents, web search, images and voice, on its own servers with any model and with roles and single sign-on. ChatGPT gives OpenAI's models with nothing to run. The open-source alternatives to ChatGPT page compares it with LibreChat, AnythingLLM, Jan and others by need.

Can Open WebUI run completely offline?

Yes. The README describes it as built to run entirely offline, with an environment variable for offline mode, and a Docker image that bundles Ollama so the model and the interface run together. Document search, local Whisper for voice and local image generation through ComfyUI keep the rest local too.

Sources

• Open WebUI README and LICENSE (Open WebUI License): github.com/open-webui/open-webui, read 2026-09-17.