Open WebUI Review 2026: Still Free to Self-Host — With a License Clause Worth Actually Reading
Open WebUI is the interface most people mean when they say "I put a ChatGPT-style UI on my local models." That's still true, and it's still free for internal use of any size. What most people miss is that in April 2025 the project moved off BSD-3-Clause to a custom, non-OSI-approved license with a branding-protection clause — a change that sparked real community backlash and forked at least one Hacker News thread. This review covers what's actually free, what the Enterprise License adds, and whether the license change should change your decision to adopt it.
| Self-hosted, standard branding | Free, any deployment size |
| License | Custom, non-OSI (Apr 2025+) |
| Remove branding at 50+ users | Requires Enterprise License |
| GitHub stars (live) | 152,720 |
| Backend it's built for | Ollama + any OpenAI-compatible API |
Open WebUI is still genuinely free. "Open source" is now a more complicated word for it.
"You can self-host Open WebUI today, for free, for a team of any size, and nothing changes for you as long as you keep the 'Open WebUI' name visible. What changed in April 2025 is that the license stopped being OSI-approved open source: a branding-protection clause now requires an enterprise license (or ≤50 users) to remove that branding, and new contributions require signing a CLA. The maintainers' stated reason — stopping rebrand-and-resell operators from free-riding on their support burden — is defensible. Whether you agree the trade-off was worth losing the OSI label is a real debate, not a manufactured one."
Feature-for-feature, Open WebUI remains the deepest self-hosted AI interface most people will ever need: RAG across 9 vector databases, voice/video calls, Notes, Channels, Calendar, persistent memory, and a plugin system that speaks MCP. The license story doesn't change any of that — it changes what "open source" means when you tell someone you're using it.
A self-hosted chat platform built to sit in front of any model backend.
Open WebUI is a single Python/Svelte application (Docker, pip, uv, or Kubernetes) that gives you a ChatGPT-style interface over local models via Ollama or any OpenAI-compatible API — LM Studio, GroqCloud, Mistral, OpenRouter, vLLM, and more, mixed and matched freely. Its own GitHub description is intentionally minimal: "User-friendly AI Interface (Supports Ollama, OpenAI API, ...)."
The one asterisk worth flagging early: as of v0.6.6 (April 19, 2025), Open WebUI's license is not OSI-approved open source — it's a custom license with a branding-protection clause, even though the maintainers still call it "very permissive."
Open WebUI's real 2026 story is community scale plus a licensing fork in the road.
open-webui/open-webui, verified on GitHub, Sep 22, 2026.
Verified on GitHub, Sep 22, 2026.
Independent paper documenting the project.
Custom "Open WebUI License."
| Era | License | What it meant |
|---|---|---|
| Through v0.6.5 | BSD-3-Clause | Fully OSI-approved open source; anyone could rebrand and resell freely |
| v0.6.6+ (Apr 19, 2025) | Custom "Open WebUI License" | Adds a branding-protection clause + CLA for new contributions; no longer OSI-approved. Sparked visible community backlash, including a top Hacker News discussion thread. |
| 2026 | Same custom license | Branding required only at 50+ users unless you hold an enterprise license or are a substantive contributor with written permission |
The praise is about feature depth. The debate is about the license.
"Open WebUI is the best all-rounder with the largest community and the broadest feature set, from RAG and web search to voice, image generation, and Python function calling" — and "the pick for most self-hosted local LLM users: fastest to set up and the best Ollama frontend."
The license change drew real, public pushback: commenters questioned trademarking a generic name like "Open WebUI" and argued trademark enforcement, not a license change, would have better targeted the rebrand-and-resell problem the maintainers described.
Comparisons consistently position Open WebUI's strength as breadth: "a broad model-and-tool workspace" versus AnythingLLM's document-RAG focus or LibreChat's multi-provider team workbench.
There's no Cloud plan to buy. The only real choice is self-host free, or license an enterprise deployment.
Free, unbranded OK
Full rebrand allowed under 50 users/30-day window.
- All features, no limits
- Fork from v0.6.5 for zero-restriction BSD-3
Free, branding required
Same features; must keep "Open WebUI" name/logo visible.
- No fee for standard-branded use
- Costs nothing until you want to remove branding
Enterprise License
Contact sales; no published price list.
- White-label / remove branding
- Priority support, feature requests
Source: docs.openwebui.com/license and docs.openwebui.com/enterprise, verified Sep 22, 2026. No official published dollar figures were found for Enterprise tiers.
Your answer depends on whether you need to remove the branding.
Self-host for free, rebrand freely if you want — no restriction applies.
Still completely free — the branding clause only restricts removing the name.
You need an Enterprise License — contact sales for a custom quote.
Fork from v0.6.5, the last pure BSD-3-Clause release, with zero restrictions.
Open WebUI is the most-cited pick for exactly this — install and go.
Notes, Channels, Calendar, and persistent memory turn it into more than a chatbot skin.
The hero screenshot above already shows the actual breadth in the sidebar: Notes (a dedicated writing workspace with AI rewrite), Workspace, Channels (real-time shared spaces where a team and AI models collaborate in one timeline), Folders, and per-model chat history — all in front of whatever backend model you've connected. The diagram below maps how those pieces relate.
- Local RAG spans 9 vector databases with hybrid BM25 + vector search and reranking.
- Persistent Memory carries facts across conversations, not just within one chat.
- Voice/video calls integrate multiple STT (Whisper, OpenAI, Deepgram, Azure) and TTS (ElevenLabs, Azure, OpenAI) providers.
- This much feature surface means a real admin/RBAC learning curve for anything beyond solo use.
- Quality of RAG/voice features still depends on which backend model and API keys you connect.
Deploy once, then point it at whatever models you want.
Code execution is sandboxed by default — a detail worth knowing before you enable it.
Open WebUI's plugin system (Filters, Actions, Pipes, Tools, Skills) lets you extend the platform without forking it, and its own Automations feature schedules prompts to run on a recurring basis, with each run linking back to the chat it produced. Per Open WebUI's own security documentation, Python code execution defaults to a sandboxed Pyodide/WASM environment — server-side execution is opt-in, not the default, which directly answers the "code execution without sandboxing" concern that automated supply-chain scanners sometimes flag.
Any OpenAI-compatible backend, plus a wide net of web-search and RAG providers.
Beyond Ollama, Open WebUI connects to LMStudio, GroqCloud, Mistral, OpenRouter, and vLLM as model backends, and to a long list of web-search providers for RAG-augmented answers — SearXNG, Google PSE, Brave Search, Kagi, Mojeek, Tavily, Perplexity, Firecrawl, Bing, Jina, Exa, Azure AI Search, and Ollama Cloud among them, per the official README.
| Category | Examples (official README) |
|---|---|
| Model backends | Ollama, LMStudio, GroqCloud, Mistral, OpenRouter, vLLM, any OpenAI-compatible API |
| Vector databases (RAG) | 9 supported, plus Tika/Docling/Document Intelligence/Mistral OCR/PaddleOCR-vl extraction |
| Web search providers | SearXNG, Google PSE, Brave, Kagi, Tavily, Perplexity, Bing, Jina, Exa, and more |
| Speech | Whisper, OpenAI, Deepgram, Azure (STT); Azure, ElevenLabs, OpenAI, Transformers, WebAPI (TTS) |
There's no Open WebUI bill for inference — you pay whatever your connected backend charges.
| What you're doing | What you pay |
|---|---|
| Self-hosted, standard branding, any user count | $0 to Open WebUI; only your infrastructure and backend model costs |
| Self-hosted + Ollama backend | $0 for local inference, or Ollama Cloud's own metered pricing if you route there |
| Self-hosted + a paid API backend | Whatever that provider (OpenAI, Groq, OpenRouter, etc.) charges directly |
| Enterprise License (white-label, 50+ users) | Custom quote; no official published price list. Third-party trackers report ~$5K/yr as an unverified estimate. |
An unusually transparent public security posture, for a project this size.
Source: docs.openwebui.com/security, official — checked 2026-09-22.
People who want one interface over every model they run.
| Situation | Why Open WebUI fits | Likely path |
|---|---|---|
| Solo developer running local Ollama models | Widely cited as the best Ollama front end; fastest setup | Self-host, standard branding |
| Small internal team (<50 users) | Free, unrestricted, can even rebrand if desired | Self-host, any branding |
| Larger internal team (50+ users), name visible is fine | Still fully free — branding clause only restricts removal | Self-host, standard branding |
| Vendor wanting to white-label for customers | Needs branding removed at scale | Enterprise License, custom quote |
| Document-Q&A-first workflow, no Docker/terminal wanted | Open WebUI is broader but more technical to deploy | Consider AnythingLLM's desktop app instead |
Four real gaps, not manufactured ones.
The v0.6.6+ branding clause means the license fails OSI's certification criteria — a real change of category, not just a cosmetic one, even though day-to-day self-hosted use is unaffected for most people.
No official dollar figures for the Enterprise License were found on docs.openwebui.com as of this check — you have to contact sales to learn the real cost.
Public discussion (including a well-known Hacker News thread) raised real objections about trademarking a generic project name and about relicensing after community contributions. This is documented in-article as a real, ongoing debate, not resolved either way.
RBAC, plugin security implications, and production-hardening are all real considerations the project's own docs dedicate entire pages to — this isn't a five-minute toy deployment at team scale.
If Open WebUI's breadth or license isn't what you need, here's how to think about it.
| If you mostly need... | Compare Open WebUI with... |
|---|---|
| Document Q&A and RAG-first workflows, with a native desktop app | AnythingLLM — no Docker/terminal required |
| A multi-provider team workbench with per-user token tracking | LibreChat — LDAP/OAuth/social login, usage tracking |
| An OSI-approved permissive license with zero branding conditions | Fork from Open WebUI v0.6.5 (pure BSD-3-Clause) and build from there |
| The local model backend Open WebUI is most commonly paired with | Ollama — read our review |
No affiliate relationship shapes this review.
JAVIS found no evidence of a public Open WebUI affiliate or referral program as of this check — the project's only paid offering is direct Enterprise License sales. Every CTA here points to the official product, unmodified.
Go to the official page.
Open Open WebUIRead the alternative that matches your real question.
Choose the next articleOpen WebUI makes the most sense next to the backend it sits on top of.
Questions people are actually searching right now
Is Open WebUI free?
Yes, for self-hosted use of any size, as long as you keep the "Open WebUI" branding visible. Removing branding at 50+ users requires an Enterprise License; deployments of 50 or fewer users may rebrand freely.
Is Open WebUI still open source?
Not in the OSI-approved sense as of v0.6.6 (April 19, 2025). It moved from BSD-3-Clause to a custom "Open WebUI License" with a branding-protection clause, which the license's own FAQ states does not meet OSI's certification criteria. All code up to and including v0.6.5 remains pure BSD-3-Clause and forkable with zero restrictions.
What triggered the license change?
The maintainers say some groups were stripping branding, reselling the software as their own, and then relying on the original team for free support — straining a small independent team. The change added a branding requirement rather than closing the source or adding usage fees.
How much does the Enterprise License cost?
Open WebUI does not publish Enterprise pricing; you have to contact sales for a custom quote. Third-party pricing trackers report figures like ~$5K/year for 50+ users, but this is not officially confirmed and should be verified directly.
Does Open WebUI work with Ollama?
Yes — it's widely cited as the most common and best-supported front end for a self-hosted Ollama server, connecting via Ollama's OpenAI-compatible REST API.
Is Open WebUI better than LibreChat or AnythingLLM?
Different strengths: Open WebUI wins on overall feature breadth and community size; AnythingLLM wins for document-RAG-first workflows with a native desktop app; LibreChat wins for multi-provider team chat with per-user token tracking.
Is Open WebUI's code execution safe?
By default, yes — Python code execution runs sandboxed via Pyodide/WASM; unsandboxed server-side execution is opt-in, not automatic, per the project's own security documentation.
Can I use Open WebUI without Docker?
Yes — it also installs via pip, uv, or Kubernetes (kubectl/kustomize/helm), though Docker with the `:ollama` or `:cuda` tags is the most common path.
Where this came from, and when it was checked.
GitHub stats (stars/forks/license): GitHub data for github.com/open-webui/open-webui, Sep 22, 2026.
License history and terms: docs.openwebui.com/license, official, read in full, Sep 22, 2026.
Enterprise licensing page: docs.openwebui.com/enterprise, official; no published dollar figures found on the official page as of this check.
Security posture: docs.openwebui.com/security, official, covering Security Policy, Supply Chain scanners, Chat Data Privacy, Vendor Dispositions, and Accepted Risks.
License-change community reaction: Hacker News item 43901575, a public discussion thread, corroborated by independent tech-press coverage of the backlash.
Real screenshots/media: the official README demo.png and the GitHub-rendered repository card (opengraph.githubassets.com).
Feature list: the official open-webui/open-webui README, read directly.
Academic citation: arXiv 2510.02546, an independent paper documenting Open WebUI, referenced as a market/legitimacy signal, not summarized for technical claims.
User sentiment: independent self-hosted-AI comparison pieces (Open WebUI vs. LibreChat vs. AnythingLLM); Hacker News community reaction to the license change.
