Use OpenAI or switch to Claude, Gemini, Mistral or a local model, in a workspace you run. ChatGPT Teams has the better chat experience; this page is about everything around it.
Book a demo Try the live demoWith the guard service installed and placeholders switched on (from Pro), the model gets a tag instead of the name. You read the answer with the name restored.
Write a polite reply to Anna de Vries about her late invoice.
Write a polite reply to [PERSON_1] about her late invoice.
The model answers: Dear [PERSON_1], thank you for your message. We are looking into your invoice.
You read: Dear Anna de Vries, thank you for your message. We are looking into your invoice.
Based on public sources; check the vendor's own documentation.
| Bee Flow | ChatGPT Teams | |
|---|---|---|
| Chat experience | Good and improving, with voice chat from Starter, but a company whose whole product is the chat keeps an edge | The better chat interface, with new model capabilities first |
| Model choice | Eight hosted providers, local runtimes or any OpenAI-compatible endpoint | Switching models means switching products |
| Who operates the workspace | You, on your own infrastructure | The model vendor holds the conversations, runs retrieval and decides the region |
| Personal data and the provider | With the guard service installed: detected on your own hardware and blocked before the prompt leaves, from Free. From Pro it can be swapped for placeholders that are restored in the answer | A contractual assurance rather than a log entry |
| Around the chat | Knowledge bases, automations and all built-in integrations from Free; meeting notes, team projects and a support inbox from Pro | Built around the chat window |
Encryption with managed keys is included from Pro. In zero-knowledge mode (Enterprise), private chat and notebook conversations are encrypted with a key derived from the user's password. Knowledge-base documents and meeting transcripts stay readable by the server, because search and transcription must read them.
A self-hosted install checks its licence key offline and reaches no Bee Flow service unless you switch that on, so an air-gapped install keeps running. The workspace and its stored conversations stay on infrastructure you control; what leaves is what you send out yourself, such as prompts to a hosted model or calls to an integration you connect.
You want the best chat interface with the newest model capabilities, and your organisation has no constraint on where conversations are processed. OpenAI ships to its own product first, and the apps and voice experience are excellent.
You need to show whether personal data was sent to a provider, or you want knowledge bases, automations and meeting notes around the chat. Chat, knowledge bases and automations are already in the Free licence; meeting notes come with Pro.
Yes. Bee Flow has an OpenAI adapter and an Azure OpenAI adapter. You bring your own API key. The conversation history and retrieval live in your workspace instead of the provider's.
That is what the Privacy Shield is for. With the guard service installed it checks prompts on the way out and blocks personal data, in every licence including Free. From Pro it can swap that data for placeholders that are restored in the answer, add your own data types and hold personal data back from web searches.
No. Besides self-hosting there is an EU cloud option; the page Self-host or EU cloud compares the two. Self-hosting is what makes "on your own infrastructure" in the table above literally true, but it is not required to use the product.
They stay where they are: there is no import for another vendor's chat history. Most teams start fresh, because the conversations worth keeping usually produced documents, and those you can upload into a knowledge base on day one.
Bring your API key, send a prompt with a sample customer name in it, and look at what the provider actually received.
Book a demo Install free