Pricing
Self-hosting costs nothing and has no caps. The hosted version is a subscription, because the servers and the model calls are ours to pay for.
Run it yourself How the hosted version worksPrices, limits and trials come straight from our billing configuration, so what you see here is what checkout charges.
Read this before comparing us to a per-seat tool
On a hosted plan the price is also your monthly AI budget: the plan amount is the ceiling on what your organisation can spend on model calls that month. You are told at 80% and stopped at 100%, rather than discovering it on an invoice. Where a hosted plan carries seat or source limits, they are printed on the plan card itself — and the section above always reflects what checkout would actually charge, because both read the same billing configuration. Nothing here is metered per message, and there is no overage bill.
A subscription
We run it on Scaleway in EU data centres, keep it patched, and handle backups and TLS. You get billing, the compliance hub, the website builder and its analytics, and a Dutch legal-sources knowledge base we maintain. A plan sets your seat count, assistant count, knowledge sources and monthly AI budget — and the cards above come straight from the live billing configuration, so what you see is what exists. Not seeing a plan that fits? Talk to us about a hosted setup before assuming the answer is no.
Free, or a licence key
Community costs nothing, needs no licence key, and enforces no caps at all - no seat limit, no message limit, no spend ceiling, because the model bill is yours. Specialist modules (meeting notes, notebooks, App Studio, the compliance hub, the support inbox) need an Enterprise licence key, which unlocks them without moving your data anywhere and is priced per deployment rather than per seat - talk to us. Licence checks are offline against a bundled public key, so an air-gapped install keeps working. The editions page lists the two side by side, row by row.
FAQ
Yes, and that is a property of the code rather than a promise: the usage-limit checks return immediately unless the deployment is our cloud. On your own server there is no seat cap, no message cap and no spend ceiling to hit — and all 44 built-in integrations are in Community, not held back for a higher tier.
Nextcloud sign-in ships in Community. Google and Microsoft SSO — including Entra ID group sync — are part of Enterprise. The identity page maps out exactly what each provider brings and what is deliberately not implemented.
Organisation admins get a warning email at 80% of the monthly AI budget, and requests stop at 100% with a plain message rather than a silent failure. Seat, assistant and knowledge-source limits refuse the action that would cross them and say which limit was reached.
No. The free tier involves no checkout at all, and where a trial is offered it is created without collecting a payment method - at the end it cancels rather than charging you.
Yes, in both directions. It is the same codebase, and websites, knowledge bases and automations all export to files. What changes is who runs the servers and who pays the model bill.
Source-available, and we would rather be precise than borrow the word. The server and frontend are published under the Sustainable Use License v1.0: read it, modify it, run it internally without paying — what you may not do is resell Bee Flow as your own hosted service. It is not an OSI-approved open-source licence. We are working toward AGPL-3.0-or-later; today the Nextcloud connector is the only part that carries it.
Because model calls are the real cost of running this, and we would rather cap them visibly than meter you by the message and send a surprise invoice. If your team needs more headroom, the answer is a larger plan, not an overage.
Describe the team and where the data has to live. If self-hosting Community already covers you, we will say so.
Compare the editions Read the self-hosting guide