Open WebUI

open-webui/open-webuiReviewed by AI

What it does

Open WebUI is a chat interface you install on your own computer or server and then open in a browser. It talks to models running locally through Ollama and to any OpenAI-compatible service, so your conversations and uploaded documents stay on the machine you run it on. Besides chat it has document search, web search, notes, a calendar, voice calls, image generation and user accounts with permissions. You install it with one command using Python (pip) or Docker, and you still need either a local model or an API key to get answers.

How easy it is to install

C · install 1 or 2 tools · Windows

Windows C
macOS C
Linux C

There is no installer: on all three systems you install Python 3.12 and run `pip install open-webui` then `open-webui serve`, or use the one-line Docker command the README gives. After that you open http://localhost:8080 in a browser, create the first account and point it at Ollama or paste an API key.

Numbers on GitHub

Stars: 157,407on Oct 31, 2026
Downloads: —all releases
Growth: +550stars in 5 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Python, Docker
To build from source: Python, Node.js, Docker
Disk space: 250 MB–950 MB · estimate
Accounts or keys: OpenAI
Cost: Freemium · Free to install and run yourself. The project also sells an Enterprise Plan, and the license only lets you change or remove the "Open WebUI" branding below 50 users unless you buy an enterprise license. Any paid AI service you connect is billed separately.
License: BSD-3-Clause (Permissive)
Can replace: ChatGPT Plus, TypingMind, NotebookLM
Last update: Oct 29, 2026

Risk signals (6)

Attentionlicense forbids removing branding
The LICENSE file is not plain BSD: clause 4 says licensees "are strictly prohibited from altering, removing, obscuring, or replacing any 'Open WebUI' branding" except for deployments of at most "fifty (50)" users in a 30-day window, with written permission, or with an enterprise license. GitHub reports the license as NOASSERTION and pyproject.toml classifies it as "License :: Other/Proprietary License".
AI review
Attentioninsecure network defaults in example
.env.example ships `CORS_ALLOW_ORIGIN='*'` and `FORWARDED_ALLOW_IPS='*'`, with the file's own warning: "WARNING: * trusts forwarded headers from every connection. Use only behind a trusted proxy". If you put this on the open internet with the example settings as written, that warning applies to you.
AI review
Attentionplugins install their own code
.env.example describes a "Master switch for internal Tools/Functions and external OpenAPI/MCP/Open Terminal plugins" with `ENABLE_PLUGINS=true` by default, and `ENABLE_TOOLS` controls "workspace Tools and their dependency installation". Tools you add from the community can therefore pull in and run extra code on your machine.
AI review
Attentionoptional ai terminal access
README: Open Terminal "Give your agents a terminal and filesystem to carry out multi-step tasks. Let them analyze data, run scripts, fix errors". This companion piece is optional, but once connected the AI can run commands on the machine hosting it.
AI review
Infocontributor agreement required
LICENSE: "By contributing to this project, you agree to the project's Contributor License Agreement", and the repository has a CONTRIBUTOR_LICENSE_AGREEMENT file plus a LICENSE_HISTORY recording that earlier code was under different terms.
AI review
Infotelemetry off in shipped settings
Both .env.example and the Dockerfile set `SCARF_NO_ANALYTICS=true`, `DO_NOT_TRACK=true` and `ANONYMIZED_TELEMETRY=false`, and the Dockerfile writes an all-zero id to the bundled vector database's telemetry file.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether it actually installs and runs on each system, since Radar did not run it
Real disk use, which grows a lot once the speech and document-search models are downloaded on first use (the Dockerfile pulls Whisper and sentence-transformers models)
What a community Tool, Function or MCP server does once you install it, and what extra packages it pulls in
The real monthly cost, which depends on whatever AI service, web search or storage provider you connect
Whether your installation ends up reachable from the internet, which depends on how you set up Docker, ports and any proxy