Credits

What a credit is, before you buy one

A credit is prepaid usage of our managed model route — Aerenium sends the request, we carry it to the model and bill the tokens it used. No more than that. It is not a subscription, it does not expire, and it is not one flat rate stretched over every model. Buy a balance, spend it as you use models, top it up when you want to. (Looking for encrypted backup? That is a separate product: Cloud.)

Spent only when a request runs Nothing is drawn down for holding a balance, and there is no monthly minimum or auto-renewal.

Never expires A balance bought today is still whole in two years if you never touch it.

Priced per model Every model carries its own rate, read live from the catalog, so nothing quoted goes stale.

The arithmetic

How a request becomes a number

Billing is per-model-exact. There is no flat token rate, no rounding everything into three price bands, and no fixed multiplier applied on top of a guess. A call is priced from what it actually used:

Input tokens — everything sent up: your message, the conversation so far, and any instructions Aerenium adds behind it. Long conversations cost more for this reason alone.
Output tokens — everything the model generates. Output is usually the more expensive direction, frequently by several times over, which is why the two are priced separately.
Images and per-call fees — if a model charges for image inputs or a flat fee per request, that is billed too, at the rate the model declares.
Free models — models that cost nothing upstream bill only a small flat processing fee.
Our margin — charged once, in the pack price, and printed on every card on the packs & rates page. Model rates are passed through at raw catalog cost; spending adds nothing on top.

Model choice moves your spending far more than pack size does. A reasoning-heavy frontier model can cost dozens of times what a small fast model costs for the identical exchange — which is why rates are published per model rather than averaged into bands.

Mechanism

How a request travels

1
Your request leaves AereniumIt travels to Aerethis signed with the credential you were issued when you logged in. There is no anonymous route through the proxy, and no way to configure one.
2
We attach our provider credentialsYour keys are never involved, because there are none. The provider sees us, not you.
3
Failures are handled before you see themIf a route errors or times out, the request retries and fails over to another provider serving the same model, rather than surfacing to you as a broken reply.
4
The reply returns and your balance is drawn downYou are charged for what the request actually used — input and output counted separately, at that model's raw catalog rate, exactly as the packs & rates page shows it. Our margin was already in the pack price; spending adds nothing.

Data handling

What we log, and what we never log

Recorded on every request

Which model you called
How many tokens went in, and came out
Credits charged
How long it took, and whether it succeeded
Timestamp, and which account it belongs to

That is what a bill is made of, and it is all we keep.

Never recorded

Your prompts
The model's replies
Files, screen captures, or anything else Aerenium reads from your machine

We do not train on your content, and we do not sell it.

We can only speak for our half — so here is the honest rule.

Our layer adds real privacy: no prompt logging, no training, no resale. But every prompt still reaches the company running the model, because a model cannot answer what it cannot read — and what they keep is governed by their policies, not ours. So treat every AI, ours included, like a stranger on the phone: never send passwords, ID numbers, medical or financial records, or anything that could identify a person that you would not hand to an outside vendor. For sensitive work use your own key with a provider you trust, or a local model — nothing beats a request that never leaves your machine. Full privacy guide →

Throughput

What your plan changes

Credits work on every plan, including Free. What the plan changes is how fast you can send — nothing is metered by request count, so the cap itself never costs you anything.

Plan	Requests per minute
Free	10
Plus	60
Pro	300
Pro Max	600
Enterprise	1,200

The limit rolls continuously, and exists so one account cannot starve another. Subscribers also earn bonus credits on every pack — see the packs & rates page.

If you stop using it

What happens to credits you have not spent

Credits do not expire and uninstalling the app does not consume them — a balance lives on our side, attached to whatever you bought it with. What matters is whether that something is recoverable.

With a full account — your balance is tied to your email address. Uninstall, reinstall, change computers: sign in and it is still there. Nothing is lost by leaving.
With a token account — the 16-character token is the only thing identifying the balance, and it is stored on your machine. If you uninstall without copying it, the credits become unreachable: there is no email, no name and no payment record we can match you against, so we cannot restore access. Copy the token before you go.
Dormant token balances — a token account with no balance and no activity for 90 days is removed, and a token balance untouched for ten years is written off. Both are housekeeping for abandoned records rather than an expiry date on money you are using.
Refunds — credits are sold as a prepaid balance and are not refundable once purchased, which is also why nothing forces you to spend them on a schedule.