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

# Get your AI budget and gateway controls

> This month's model tokens against your organisation's AI budget, and whether prompt redaction, the answer cache, fallbacks and canaries are on.

Returns your organisation's monthly AI budget against the model tokens it has used this calendar month (UTC), and whether your deployment redacts personal data from prompts, caches answers, and routes models to fallbacks or a canary. There are no parameters: the organisation is the one your credential belongs to. See [AI gateway controls](https://docs.mithunai.com/gateway/answer-controls.md).

Every control is set by whoever runs your deployment. There is no route to change one.

When `budget.mode` is `hard` and `budget.percent_used` reaches `100`, every request that would call a model is refused with `429` and the code `budget_exceeded` until the month turns. See [Errors](https://docs.mithunai.com/api-reference/errors.md).

The response carries tokens and percentages only. A cost budget shows as `cost_limit_set` and in `percent_used`, never as an amount.

Requires the owner, admin or editor role. Members and dataset operators receive `403`.

## Response

**period** `object` *required*

The calendar month. `end` is when the budget starts again.

**budget** `object` *required*

**properties**

**enabled** `boolean` *required*

Whether any budget applies.

**mode** `string` *required*

`soft` only emails your owners and administrators at 50, 80 and 100 per cent. `hard` also
stops model calls at 100 per cent.

**token\_limit** `integer | null` *required*

Model tokens (input plus output) a month. `null` is no token budget.

**cost\_limit\_set** `boolean` *required*

Whether an estimated-cost budget also applies.

**tokens\_used** `integer` *required*

Model tokens used this month, on every surface together.

**percent\_used** `integer | null` *required*

Of the tighter of the two limits, rounded down. `null` with no budget.

**alert\_thresholds** `integer[]` *required*

Always `[50, 80, 100]`.

**redaction** `object` *required*

`enabled`: whether personal data is replaced with placeholders before a prompt leaves.

**cache** `object` *required*

`enabled` and `ttl_seconds`: whether a repeated question is answered from the cache, and for how
long an answer is kept.

**routes** `object[]` *required*

One entry per model with a fallback chain or a canary: `model`, `fallbacks` in order, and `canary`
(`model` and `percent`) or `null`.

**managed\_by** `string` *required*

Always `deployment`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.