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

# Credit Lifecycle

> Understand Starfire AI included, bonus, purchased, organization, period-based, and hard-ceiling credit behavior.

# Credit lifecycle

Credits measure eligible Starfire compute and feature usage. The available balance can come from more than one source and can be governed by billing periods and hard limits.

## Credit sources

Depending on plan and active product configuration, a billing context can contain concepts such as:

* plan-included credits
* organization shared credits
* administrative bonus credits
* purchased credits when supported
* temporary promotional or recovery credits

## Usage attribution

Every metered workload should be attributable to the context that owns it:

* personal Chat
* organization Project
* FORGE build
* Research run
* developer API application
* agent/workflow run

## Billing periods

Plan-included allowances can reset or renew according to the subscription/billing period. The active billing surface is the source of truth for current period dates and available allowance.

## Hard ceilings

A hard credit ceiling blocks additional metered work after the configured threshold even if an upstream provider would still accept the request.

This lets users and organizations bound spend or usage.

## Bonus adjustments

Authorized administrators can apply audited credit adjustments when product policy allows it. Bonus credits should be visible as a distinct source rather than silently rewriting historical usage.

## Reconciliation

If renewal occurred but the expected allowance did not appear, troubleshoot period/provisioning state before manually adding credits.

## Shared organization credits

Shared pools should still preserve attribution by member/application/workload. A shared balance should not make it impossible to discover what consumed the credits.

<Note>
  Starfire does not publish one permanent credits-per-token conversion in the docs because model/provider cost and platform metering configuration can change independently.
</Note>