# Usage and limits

How many AI resources have been spent, how much of the Bitrix24 account monthly quota is already used, and during which hours the quota drains slower.

Usage statistics are counted **per API key**: tokens, cost and a per-model breakdown for the selected period. The quota is counted **per Bitrix24 account**: a single monthly limit shared by all of its keys. Off-peak is the discount schedule that makes that quota drain slower during certain hours of the week.

Scope: `vibe:ai`

## Operations

- [Usage statistics](./consumption/usage.md) — `GET /v1/ai/usage`
- [Bitrix24 account AI quota](./consumption/quota.md) — `GET /v1/ai/quota`
- [Off-peak schedule](./consumption/off-peak.md) — `GET /v1/off-peak`

## Typical scenario

1. Check the remaining quota before bulk processing: [`GET /v1/ai/quota`](./consumption/quota.md). The `pctUsed` and `exhausted` fields let you warn the user before a `402` is returned.
2. Find out when the quota drains slower: [`GET /v1/off-peak`](./consumption/off-peak.md), and move deferrable tasks into the discounted windows.
3. After processing, check the spend for the key: [`GET /v1/ai/usage`](./consumption/usage.md).

## See also

- [AI Router](/docs/ai)
- [Model list](/docs/ai/models/list)
- [Request limits](/docs/ai/chat/rate-limits)