A license we agree with you. AI billed as you use it.
Geist is in private beta, and we set the license with each organization. AI usage follows one model: the provider's list rate for what you use, plus a fee for providing model access.
The license and the AI you use
Talk to us
Access to Blueprint, Finder, Horizon and Wisp for your organization. We agree the scope and terms with you.
Talk to usPay for what you use
Each AI call is priced at the provider's published list rate for its model, plus a fee. Calls rejected before they reach the model cost nothing.
How AI usage is billedYour keys or ours
You choose per organization. The model stays the same either way. What changes is who bills you for the provider cost.
We hold the provider accounts
Geist calls the provider with its own keys. You pay the list rate for your usage plus the fee, on one bill from us.
You bring a provider key
Admins add the key in Connections & Keys. The provider bills you directly under your own contract. We bill the fee, calculated on the list rate for the same usage.
Stopped before the model: nothing to pay
A call stopped before it reaches the model, for example by a permission check or a spend limit, processes no tokens. It costs nothing and does not count against your limits.
Why there is a fee
We run the model access for you. We route each request to the right model, keep the model catalog current, meter usage, enforce spend limits and keep the audit record. The fee pays for that work, and it applies the same way with your keys or ours.
Billing questions
How is AI usage billed?
By usage. Each call is priced at the provider's published list rate for the model it used, plus our fee. Billing terms are agreed with your organization as part of the license.
Can we use our own provider keys?
Yes. Admins add keys for OpenAI, Anthropic, Google Gemini and the other supported providers in Connections & Keys. Keys are stored encrypted per organization and can be tested, rotated and revoked. The provider then bills you directly, and we bill the fee.
What counts as usage?
The tokens a model processes for your organization: input, output, cached and reasoning tokens, at the list rate for that model. Calls rejected before inference count as zero.
Where do we see what we spend?
In the admin console under AI Spend. It shows provider cost per currency, the split into input, output, cache and reasoning tokens, which product and resource caused it, and a preview of the fee. Spend limits per provider, organization and user are set there too.
Does the fee follow our negotiated provider rates?
No. The fee is calculated on the provider's published list rate, so discounts you agree with a provider do not change it.
What does the license cost?
We agree it with each organization. Tell us about your teams and processes, and we make you an offer.
Talk to us about your setup
Tell us how many teams and processes you want to start with, and whether you bring your own keys.