> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orbitrage.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Plans & billing

> Credits, plans, and how spend is metered.

Orbitrage bills on **pre-paid credits**. You're charged per call for the
open-weight models we serve; frontier models are [BYOK](/concepts/byok) and cost
**\$0** against your credits because your own provider bills you directly.

## Plans

<CardGroup cols={2}>
  <Card title="Free" icon="gift">
    **\$5 of credits** to start. Up to **3 projects** and **100 nodes** per
    workflow. No card required.
  </Card>

  <Card title="Pro" icon="rocket">
    **$99 / month**, including **$99 of credits** per month and higher limits.
    Upgrade from the Billing page.
  </Card>
</CardGroup>

When a Pro plan lapses, the organization reverts to free-tier limits until
renewed.

## Credits & metering

Each call records a `cost_usd` — the upstream provider price plus a small
platform infra fee (currently **2.5%**). That amount is debited from your
organization's balance. The Billing page shows:

* **Credits** — total available.
* **Spent** — cumulative usage.
* **Remaining** — credits minus spent.

### What credits do and don't cover

| Charge                                                                           | Debits credits?                                     |
| -------------------------------------------------------------------------------- | --------------------------------------------------- |
| Open-weight chat models we serve (Bedrock, Fireworks, Cerebras, open Foundry)    | Yes — upstream + 2.5%                               |
| Images (`gpt-image-2`) and audio (Deepgram `nova-*` / `aura-*`)                  | Yes — per image / per minute / per 1k chars, + 2.5% |
| [Managed Tools Gateway](/concepts/tools-gateway) calls                           | Yes — they run on **our** pooled tool keys          |
| Frontier chat models on your own key (`claude-*`, `gpt-*`, `gemini-*`, `grok-*`) | **No** — `cost_usd = 0`, your provider bills you    |

<Note>
  Both rules apply within a single turn. A BYOK turn that invoked two managed tools
  is billed exactly the price of those two tool calls and nothing for the LLM
  tokens.
</Note>

When credits run out on the free tier, calls return **HTTP 402** until you top
up — so a runaway loop can't quietly drain you past your balance.

```json theme={null}
// 402 response when credits are exhausted
{
  "error": {
    "type": "billing_limit_error",
    "code": "credits_exhausted",
    "message": "Credit balance exhausted. Top up via the Billing page to continue.",
    "plan": "free",
    "spent_usd": 5.0,
    "credits_usd": 5.0
  }
}
```

## Topping up

Buy credits from the [Billing page](https://app.orbitrage.ai/billing) (presets
$10 / $25 / $50 / $100, or a custom amount). Payments are processed by Razorpay;
credits reflect within about 30 seconds of a successful payment.

<Note>
  Crediting happens server-side from a signed payment webhook, so your balance
  updates reliably even if you close the checkout window.
</Note>

## Seeing where spend goes

Use the **Cost** insights (in Overview and per workflow) or
[Ask Analytics](/platform/ask-analytics) to break spend down by model,
provider, workflow, and end-user — and to see how much routing **saved** you
versus a frontier baseline. Set a [cost alert](/platform/alerts) to get
notified before a spike becomes a surprise.