Skip to main content
9 – 17 UHR +49 8031 3508270 LUITPOLDSTR. 9, 83022 ROSENHEIM
DE / EN

CompanyGPT vs. Open WebUI: licence, permissions and cost transparency

Tobias Jonas Tobias Jonas | | 7 min read

Open WebUI is one of the most widely used open-source chat interfaces for language models — around 150,000 GitHub stars (as of August 2026) speak for themselves, and the release cadence is high.

We do not build on it ourselves, but on LibreChat. This page explains factually why — and at the same time clears up a misunderstanding we keep running into in the market.

First, the misunderstanding: the 50-user limit

You often hear that Open WebUI “costs money from 50 users onwards”. That is not true, and we say so explicitly even though it runs against our own sales interest.

The 50-user threshold in the licence relates exclusively to removing or altering the branding. The software itself may be run free of charge with 5,000 users too, as long as the Open WebUI branding remains unchanged.

The licence states verbatim that licensees are prohibited from altering, removing, obscuring or replacing the “Open WebUI” branding — with three exceptions: where the number of end users does not exceed 50 within any 30-day period, where express written permission from the rights holder exists, or where an enterprise licence expressly permits it.

For most companies this is unproblematic. It becomes relevant precisely when a platform is to be rolled out under your own name and your own visual identity.

The licence situation in detail

What many are not aware of: since April 2025 Open WebUI is no longer under the MIT licence. The history is documented in the repository:

PeriodLicence
until January 2025MIT
until April 2025BSD-3-Clause
since 18 April 2025BSD-3-Clause with branding clause

Because the licence file deviates from the standard text, GitHub’s automatic licence detection no longer maps it to a standard licence and displays it as “Other”. On top of that comes a contributor licence agreement — anyone contributing code should read its grant of rights before contributing.

For comparison: LibreChat has been under the MIT licence unchanged since 2023. That is precisely why we were able to build CompanyGPT as a maintained fork, white-labelling included, with no user threshold and no additional licence agreement. The only obligation under MIT is to ship the copyright and licence notice along with it, which we of course comply with.

Worth mentioning for context: LibreChat has belonged to ClickHouse since November 2025; the founding team moved across with it, and the licence has remained unchanged in substance since then.

The enterprise gaps we encounter in projects

Open WebUI is a strong product. For enterprise use, four points from the documentation matter to us:

Cost and usage reporting is not documented. There is no page in the documentation on consumption, cost or budgets; the documentation does not describe one. In practice this is solved with an upstream gateway. LibreChat, by contrast, comes with token transactions, cost display including euro conversion and per-user budgets — and companyDASHBOARD builds on that.

Audit logging is off by default. According to the documentation, the value AUDIT_LOG_LEVEL is set to NONE. And even after it is enabled, the chat paths are excluded from logging in the default configuration; read accesses are likewise not logged by default. If you have to meet an evidence obligation, you should configure this before the rollout, not after.

Only one OIDC provider at a time. The documentation expressly points out that only one OpenID Connect provider can be configured at present — so Microsoft and Google cannot run in parallel. For groups with several identity tenants, that is a genuine issue. SAML is not described in the documentation.

No mirroring of source-system permissions. The connectors tool for knowledge bases covers an impressive breadth — 44 connectors including SharePoint, Confluence and Jira. But it works within the access rights that are set in Open WebUI; carrying the original permissions from SharePoint over into the knowledge base is not documented. In regulated environments, that is exactly the point at which a company-wide knowledge base becomes approvable, or does not.

Where Open WebUI is stronger

Fairness is part of this, so the other direction as well:

  • Adoption and community are considerably larger than with LibreChat.
  • Release pace is higher — several minor releases per quarter.
  • SCIM 2.0 for automated user provisioning is documented; we did not find that with LibreChat.
  • The breadth of knowledge-source connectors is remarkable.
  • A genuine enterprise contract model exists, with reference customers named by name.

Anyone who prioritises these points and can live with the licence situation is making a defensible decision with Open WebUI.

Comparison table

CriterionOpen WebUI (self-hosted)CompanyGPT
Licence of the baseBSD-3 with branding clause; GitHub lists it as “Other”MIT (LibreChat) as the base, plus our add-ons
White-labellingAbove 50 users only with an enterprise licence or written permissionPossible with no user threshold
OperationYou yourselfManaged by us, in the customer’s cloud tenant
Cost reportingNot documentedcompanyDASHBOARD per user, team and model
Audit loggingPresent, disabled by default, chats excluded by defaultSet up as part of the operation
SharePoint permissionsNot documentedcompanyRAG mirrors the permissions
Office integrationFile picker from OneDrive/SharePointNative add-in through Office.js in Word, Excel, PowerPoint and Outlook
Identity integrationOIDC/OAuth/LDAP/SCIM 2.0; only one OIDC provider at a timeSet up as part of the setup
Compliance supportYour own responsibilityAI training and EU AI Act assessment included

When Open WebUI is the right choice

  • You have a DevOps team that can carry the operation permanently.
  • You need no white-labelling — or you take out an enterprise licence.
  • SCIM-based user provisioning matters to you.
  • You need no mirroring of source-system permissions.
  • You solve cost reporting with your own gateway.

When CompanyGPT is the better choice

  • When the platform is to be rolled out under your name, with no user threshold.
  • When the knowledge base has to respect SharePoint permissions.
  • When you need cost and usage transparency without operating an additional gateway.
  • When AI should be available directly in Word, Excel, PowerPoint and Outlook.
  • When nobody in-house can or wants to take responsibility for running it day to day.

Conclusion

The most important sentence on this page is the one about the 50-user limit: it concerns the branding, not the use. Anyone deploying Open WebUI unchanged can do so free of charge at any scale.

The second most important is the one about how the licence has developed: since April 2025, Open WebUI is no longer a classic open-source licence. For a company committing to a base for the long term, that is a criterion — and the reason why we built CompanyGPT on the MIT-licensed LibreChat.

Whether you end up self-hosting or having it managed is a question of capacity. Both routes are legitimate, and we will tell you honestly which one suits your team.

Sources

All statements about Open WebUI are taken from the repository and the official documentation, retrieved on 23 August 2026:

  • Open WebUI LICENSE — https://github.com/open-webui/open-webui/blob/main/LICENSE — wording of the branding clause
  • Open WebUI LICENSE_HISTORY — https://github.com/open-webui/open-webui/blob/main/LICENSE_HISTORY — licence timeline
  • Open WebUI Enterprise — https://docs.openwebui.com/enterprise
  • Open WebUI — environment variables — https://docs.openwebui.com/reference/env-configuration — audit log defaults
  • Open WebUI — SSO — https://docs.openwebui.com/features/authentication-access/auth/sso/
  • LibreChat LICENSE (MIT) — https://github.com/danny-avila/LibreChat/blob/main/LICENSE

Note on the information given: All statements about other vendors’ products are based on their publicly available documentation as of the date stated. Vendors continuously develop their products, features and terms — the vendor’s current information is always authoritative. If any statement appears outdated or inaccurate to you, drop us a line at info@innfactory.ai; we will check and correct it promptly. This comparison is not a substitute for legal or data protection advice in an individual case.

Further reading

Tobias Jonas
Written by

Tobias Jonas

Co-CEO, M.Sc.

Tobias Jonas, M.Sc. ist Mitgründer und Co-CEO der innFactory AI Consulting GmbH. Er ist ein führender Innovator im Bereich Künstliche Intelligenz und Cloud Computing. Als Co-Founder der innFactory GmbH hat er hunderte KI- und Cloud-Projekte erfolgreich geleitet und das Unternehmen als wichtigen Akteur im deutschen IT-Sektor etabliert. Dabei ist Tobias immer am Puls der Zeit: Er erkannte früh das Potenzial von KI Agenten und veranstaltete dazu eines der ersten Meetups in Deutschland. Zudem wies er bereits im ersten Monat nach Veröffentlichung auf das MCP Protokoll hin und informierte seine Follower am Gründungstag über die Agentic AI Foundation. Neben seinen Geschäftsführerrollen engagiert sich Tobias Jonas in verschiedenen Fach- und Wirtschaftsverbänden, darunter der KI Bundesverband und der Digitalausschuss der IHK München und Oberbayern, und leitet praxisorientierte KI- und Cloudprojekte an der Technischen Hochschule Rosenheim. Als Keynote Speaker teilt er seine Expertise zu KI und vermittelt komplexe technologische Konzepte verständlich.

LinkedIn