What it is
Open WebUI is a web app you run on your own server to chat with language models. It connects to local models through Ollama and to any OpenAI-compatible API, so one interface can front both local and cloud models.
It has grown well beyond a chat window: document search (RAG, retrieval-augmented generation) over a choice of nine vector databases, user groups, LDAP, OAuth and SCIM 2.0 provisioning, usage analytics, Python tools and functions, skills, and MCP and OpenAPI tool servers. Version 0.12.0, released on 10 October 2026, added two-factor sign-in, nested groups, shared chats that several people can reply in, and realtime voice calls.
Since v0.6.6 in April 2025 the project uses its own licence: BSD-3 plus a clause that forbids removing or altering the Open WebUI branding once a deployment passes 50 users in any 30-day period, unless you hold an enterprise licence. The project itself says this is not an OSI-approved open-source licence. Code up to v0.6.5 remains plain BSD-3 and can be forked.
Our review
Open WebUI is the strongest free option for putting one private chat interface in front of local and cloud models, and it is also a large web application that needs patching like one. The project's GitHub security page lists 159 advisories published in 2026, 66 rated high and one critical: an LDAP bypass that accepted empty passwords. Most are fixed quickly and disclosed in the open, which is to its credit, but an instance left on an old tag behind a public URL is a real exposure.
The admin side is what earns its place. Groups can now nest and inherit permissions, models and knowledge bases carry per-group access, SCIM 2.0 provisions users from Okta or Entra ID, and v0.12.0 added mandatory two-factor sign-in. Redis-backed sessions let it run on several workers behind a load balancer. The cost is surface area: the release notes for a single minor version run to dozens of features, and custom Python functions should be retested after each upgrade.
Operations need some care. The quick-start notes that the :main Docker tag tracks the main branch rather than the last stable release, so production should pin a version tag. Multi-replica deployments must run database migrations on a single worker first or risk corrupting the database, and a changed WEBUI_SECRET_KEY signs every user out.
The licence decides the business question. Internal use at any headcount is free provided the logo and name stay; a consultancy that wants to hand clients a rebranded portal for more than 50 users needs an enterprise licence, quoted per seat with no public price. LibreChat, under MIT, carries no such clause, which is why agencies tend to start there even though its admin tooling is thinner.
Where it earns its place
A platform engineer at a 200-person firm with one GPU server running Ollama
Give staff a private ChatGPT-style portal that mixes local models with an OpenAI key
Run the container on a pinned version tag, connect Ollama and the API under Admin Connections, and sign users in through OAuth or LDAP. The software costs nothing; the spend is the GPU box and per-token API usage, and someone must own monthly upgrades.
An IT lead at a 60-person law firm
Let lawyers search internal precedents without sending documents to a consumer chatbot
Load documents into per-group knowledge bases backed by pgvector, restrict each to the relevant practice group, and turn on mandatory two-factor sign-in. Documents stay on the firm's own volume; compliance with any framework remains the firm's responsibility, as the enterprise page says.
A managed-services agency reselling AI chat to clients
Offer each client a branded chat portal
Below 50 users per deployment the branding can be changed freely; above that, white-labelling needs an enterprise licence priced per seat on request. Factor that unknown cost in, or start from an MIT-licensed alternative.
Before you commit
- Setup
- One docker run command serves it on port 3000, and the first account created becomes admin; pip install works on Python 3.11 or 3.12, not 3.13. The full image is about 1.62 GB, the slim one about 197 MB without local embeddings or document extraction.
- Integrations
- Ollama and any OpenAI-compatible API; ChromaDB, pgvector, Qdrant, Milvus and five other vector stores; LDAP, OAuth, trusted-header SSO and SCIM 2.0; MCP and OpenAPI tool servers; OpenTelemetry traces and metrics; Google Drive and OneDrive file pickers.
- Data and privacy
- Chats, files and users live in the mounted data volume on your server; the official Docker image sets Scarf and Chroma telemetry off. No SOC 2 or ISO certification is claimed: the enterprise page says compliance responsibility sits with the deploying organisation.
- Cost at scale
- Free at any user count with branding intact; the real costs are the host (the slim image idles near 300 MiB, local models need their own GPU) and model tokens. Rebranding above 50 users and per-user Terminals containers need an enterprise licence quoted per seat.
- Limits
- The :main tag is not a stable release, multi-replica upgrades must migrate on one worker first, and the 2026 advisory count means falling behind on releases carries real risk. Python 3.13 is not supported.
What's good
- Nested groups, SCIM 2.0, LDAP and two-factor sign-in built in
- One interface over Ollama and any OpenAI-compatible API
- Choice of nine vector databases for document search
- Slim image idles at about 300 MiB of RAM
What's not
- 159 security advisories in 2026; slow patching is risky
- Branding clause blocks white-labelling above 50 users without a paid licence
- Enterprise pricing and SLA terms not published
- Fast release cadence means retesting custom functions
Best for
Teams that want a private, multi-user chat portal over local and cloud models.
Alternatives
Open WebUI: common questions
- Is Open WebUI worth it?
- A self-hosted chat portal with admin controls most paid products lack, from a project that published 159 security advisories in 2026 alone. We rate it 4.6 out of 5.
- How much does Open WebUI cost?
- Free to self-host at any scale with the branding intact. White-labelling above 50 users, and the Terminals add-on, need an enterprise licence priced per seat; prices are not published.
- What are the downsides of Open WebUI?
- 159 security advisories in 2026; slow patching is risky; branding clause blocks white-labelling above 50 users without a paid licence; enterprise pricing and SLA terms not published; fast release cadence means retesting custom functions.
- Who is Open WebUI best for?
- Teams that want a private, multi-user chat portal over local and cloud models.
- How much does Open WebUI cost for a team or in production?
- Free at any user count with branding intact; the real costs are the host (the slim image idles near 300 MiB, local models need their own GPU) and model tokens. Rebranding above 50 users and per-user Terminals containers need an enterprise licence quoted per seat.
- Is Open WebUI safe for company data?
- Chats, files and users live in the mounted data volume on your server; the official Docker image sets Scarf and Chroma telemetry off. No SOC 2 or ISO certification is claimed: the enterprise page says compliance responsibility sits with the deploying organisation.
- What are the best Open WebUI alternatives?
- The closest alternatives we have reviewed are LibreChat, AnythingLLM, LobeHub (formerly LobeChat) and Jan.
Answers checked Oct 23, 2026.
Make Open WebUI? Show this rating on your site
