> ## Documentation Index
> Fetch the complete documentation index at: https://docs.viorant.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Credits & the Viorant provider
> Run models without bringing a key — how the credit wallet works and what spends it.
You don't need an API key to start using Viorant. Every signed-in account gets a
**credit wallet** and access to the **Viorant managed provider**, so you can write a
prompt, run an agent, and see real output on your first session.
Credits are the zero-setup path. [Bringing your own key](/providers/byok) is still
fully supported and has no credit cost — use it when you want your own account,
your own rate limits, or a model Viorant doesn't manage.
## Your wallet
New accounts are granted **2,500 credits**, once. There's no card, no subscription,
and no monthly reset.
The wallet is a single balance. Everything you do that calls a managed model draws
from the same pool — there's no separate allowance per feature.
## What spends credits
Three things in the Hub spend credits, and only when you use the **Viorant** provider:
Every turn you take with the Assistant, including the ones that plan rather than build.
Generating a prompt, and improving an existing one.
Running a prompt, skill, or agent against a Viorant tier from a test panel.
Anything running on **your own provider key costs no credits** — that traffic goes
from your machine straight to the vendor and never touches Viorant's metering.
## Tiers
The Viorant provider offers two tiers. They differ in capability and in what they
cost you per call:
| Tier | Use it for | Relative cost |
| ---------- | --------------------------------------------------------- | ------------- |
| **Medium** | Everyday work — the default, and the right starting point | Moderate |
| **Low** | Fast, cheap iteration and simple prompts | Lowest |
Pick a tier wherever you choose a Viorant model — when you add one to a Space's model
pool, in the prompt builder when you generate, or in the
[Assistant](/concepts/assistant). A single prompt generation typically costs **a credit
or two on Low** and **under ten on Medium**, so the starting grant covers a lot of
ordinary iteration.
Cost scales with how much text goes in and comes out, not just the tier. A long
document in the context window costs more than a one-line question at the same
tier.
## Watching your balance
Your balance is visible in the Hub next to your profile, and the prompt builder shows
what a generation will draw before you run it.
For the detail, open the profile menu and choose **Usage**. You'll get your current
balance plus a day-by-day history — each date shows the balance you ended that day on,
and expands to the individual entries behind it, so you can see exactly what a
particular session cost.
## When you run out
At zero, calls to the Viorant provider stop rather than silently failing or billing
you. The Hub tells you the balance is exhausted and points you at
[adding your own key](/providers/byok).
Adding a key restores full functionality immediately — your own models have no credit
cost and no cap beyond whatever your provider account allows.
## Where your data goes
Managed-model calls are **metered server-side**, which means the request passes through
Viorant's control plane to reach the model. That's what makes zero-key usage possible.
If you would rather no prompt content passed through Viorant at all,
[use your own key](/providers/byok) — that path calls the provider directly from your
machine, and Viorant is not in the request path.
Use your own provider account instead — no credit cost, no metering.