Two ways to power the assistant
Under Settings → AI you choose how the assistant runs:
- Managed — use the platform's models with nothing to configure. Usage is metered against your plan's AI allowance — the seats you buy fund a shared pool — so costs stay predictable; when the period's allowance is spent, AI features pause until the allowance resets — with your billing cycle on a paid plan (the free trial's starter allowance is one-time).
- Bring your own (Enterprise plan) — connect your own provider key. Anthropic, OpenAI-compatible endpoints at a public https URL, Azure OpenAI and Amazon Bedrock are supported. A built-in Test connection check confirms it works before you save.
Your keys are encrypted at rest and never shown back to you or sent anywhere except your chosen provider.
Budgets and metering
On a paid plan the workspace allowance comes from the seats you've purchased and resets with your billing cycle; the free trial ships with a one-time starter allowance instead. Admins can set an optional per-user monthly cap and a per-reply output ceiling, and a usage view shows where spend is going so there are no surprises.
Language
The whole app is available in English and French. Each person picks their own language; the assistant replies in it, and notifications we send are written in the recipient's language.
Settings are admin-only. If you don't see them, you don't have the manage permissions for that area — ask a workspace admin.