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

# Cost control and budgets

> Attribute model spend and configure shared or per-user limits.

**Cost control** combines attributed spend, cost-quality coverage, budgets, and access to the
Price Book. Use it to assign a spending allowance to an organization, project, team, user, or API
key. Provider inference charges are separate from your [Aixy subscription](/administration/billing).

## Understand the spend you see

Choose a reporting period and inspect the total, forecast when available, and breakdowns by
project, team, user, or key. Read **Cost data quality** alongside the total: estimated,
provider-reported, partial, and unattributed requests have different evidence.

Open an unattributed reason to inspect matching requests in [Activity](/observe/activity).
Missing pricing is not zero spend. Set up [pricing and cost attribution](/guides/pricing) before
using reported totals to make budget decisions.

## Create a budget

1. Open **Cost control → Create budget**.
2. Choose **Scope** and its target. Organization administrators can manage all supported scopes;
   Project admins manage budgets for their assigned projects.
3. For organization or project scope, choose **Shared total** or **Each user** under **Applies to**.
4. Select **Daily**, **Weekly**, **Monthly**, or **Lifetime** as the reset interval.
5. Enter the USD limit, warning percentage, and enforcement mode, then save.

A shared budget covers the target's combined traffic. **Each user** gives each attributed user an
individual allowance under the same policy. For example, a project allowance of $20 for each user
is separate from a $200 shared project budget; both may apply to the same request.

Team budgets use the API key's selected team cost center. Use individual keys for personal
attribution; sharing a key attributes all that traffic to its recorded owner.

## Choose enforcement

| Mode | Result |
| - | - |
| Monitor | Tracks attributed usage and warning/exceeded states while allowing requests to continue. |
| Hard | Requires sufficient remaining capacity before forwarding a request. Every applicable hard budget must admit it. |

A warning percentage marks budget status. To send notifications, configure a
[budget-utilization monitor](/observe/monitors) with a notification channel.

## Prepare a hard budget

Hard admission needs a conservative maximum request cost: complete applicable prices, a supported
bounded text request, and a positive output-token limit. It reserves capacity before contacting a
provider. Prepared reference prices must be sufficiently fresh; explicit Price Book rules can
provide the required rates. Follow the [hard-budget pricing checklist](/guides/pricing#prices-and-hard-budgets).

After execution, complete accounting can settle the reservation to the attributed amount.
Incomplete usage or an uncertain upstream outcome can retain conservative capacity. The reserved
amount is not a confirmed provider bill, and a timeout does not establish zero cost. Routing can
produce several upstream attempts with separate costs.

## Verify the limit

Select the budgeted key and model in Playground, resolve any pricing notice, and make a short
request. Inspect its budget decision and cost evidence in Activity, then refresh Cost control.
Members can review applicable personal budget information in [Usage](/observe/usage).

When editing an existing budget, you can change its amount, warning threshold, and enforcement.
Choose scope, allocation, and reset interval when creating the budget. Review every applicable
limit when investigating a rejection; increasing one limit does not bypass another.


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