How billing works
Credits or your own provider keys: what each one covers, how they differ, and how to tell which one you are on.
Models cost money to run, and there are two ways that cost reaches you. An organisation is on one or the other, never both.
Asteria Cloud credits. You hold a balance with us. Models are provisioned for you, running them draws that balance down, and one invoice covers everything.
Your own keys. You connect provider accounts you already hold, on the Models page. Each provider bills you directly, on their invoice, at their prices. Asteria Labs never sees the money.
Which one you are on was settled when your organisation was set up. The quickest way to tell: open Models. If the providers are there already and you cannot edit them, you are on credits. If the page is empty until you add something, you are on your own keys.
On credits
Your organisation arrives configured. A set of models we recommend is provisioned and assigned to every job the product does: conversation, indexing documents, reranking search results, generating images. You can chat and upload a document without setting anything up.
You can still change what runs. The recommendations are a starting point, not a lock, and the console marks which models are the recommended ones so you can tell what you are moving away from. Changing the chat model is the one most organisations should actually consider, once you know how your people work.
The embedding model is the exception. Everything you have already indexed was indexed with it, so changing it re-processes your entire corpus. See Changing the embedding model before you touch that one. The others you can change freely.
Your members choose by posture, not by price. In their model picker each model we provide carries a short label, Fast, Balanced or Most capable, so choosing between two models does not require anybody to read a price card.
What you spend is what the model actually cost to run, measured per request, not a per-seat or per-task fee. A quiet week costs less than a busy one. The Credits page carries your balance, your runway, where the credits went and per-member limits.
On your own keys
You configure everything, on the Models page: which providers, which models, their context windows, their prices. Anything your provider offers is available to you, including an OpenAI-compatible endpoint you host yourself.
Your providers bill you, so the amount never passes through us. The console gives you an estimate alongside it, computed from the prices you entered, so keeping those current is what makes it a figure you can plan against. See Usage and costs.
You hold the provider relationships: the keys, the quotas and the prices are yours to set, and stay current when you update them.
Side by side
| Asteria Cloud credits | Your own keys | |
|---|---|---|
| Who invoices you | Asteria Labs, once | Each provider, separately |
| Which models | A catalogue we curate and keep current | Anything your providers offer, plus endpoints you host |
| Setting up | Done for you, with recommendations already applied | You add each provider and model, and price them |
| Changing a model | Yours to change, recommendations marked | Yours entirely |
| Cost figures here | Based on your exact usage | An estimate, from the prices you entered |
| Keys and quotas | Handled for you | Yours to manage |
| Resilience | Each model already routes across several upstream providers | As good as the chain you build |
Which one suits you
Credits if you would rather not open accounts with several model vendors, if you want one invoice, or if nobody on your side wants to own keys and quotas. We handle model provisioning and updates for you, and your only recurring task is to move to the latest chat models every once in a while, so your organisation keeps benefiting from improving model intelligence.
Your own keys if you already hold provider accounts and want the bill to keep arriving where it does now, if you want a model of your own choosing, or if you run something self-hosted. You keep full control of the model list, the provider relationships and the prices the console reports against.
What is the same either way
Every feature. Storage, collections, workspaces, agents, skills, workflows and evaluations all work identically however you pay.
Moving between the two is arranged with your Asteria Cloud representative.