AI Cost Management

Visibility and control over AI spend across your organization

AI Cost Management gives you visibility into your organization’s AI model spend. It records events for every request that passes through AI Gateway, attributes each event to a user, group, or team, prices it against the model catalog, and surfaces the data in real-time dashboards.

You can set up budgets with notification thresholds to alert your team before spend reaches a ceiling. Use attribution to break spend down by user, application, or any custom dimension that you want to define with tags.

AI Cost Management dashboard showing spend over time and a breakdown by model and user

Key references

Analytics and debugging

Explore spend on the Dashboard, break down usage by user or model in the Explorer, and inspect individual events in the Events debugger.

Budgets

Create budgets with spend ceilings and notification thresholds. Monitor how spend tracks against each budget across the organization.

Attribution

Understand how each usage event is attributed to a subject, group, and team, and how to add custom tags for dimensions your teams care about.

Price catalog

Configure model prices that AI Cost Management uses to calculate the cost of each request. Add organization-level overrides for negotiated rates.

Exports

Download usage events and FOCUS cost-and-usage data as CSV or Parquet files, either once for a specific range or on a repeating schedule.

API reference

Browse the full REST API for AI Cost Management: list and query usage events, manage exports, budgets, notification channels, and the price catalog.

Frequently asked questions

No. Budgets track spend and send notifications when thresholds are crossed, but they don’t block or rate-limit requests. Budget enforcement is planned for a future release.

The model had no price configured when the request ran. An omitted cost and a reported zero cost are stored the same way. To price future requests, add the model to the price catalog or set cost fields on the AI Model target in AI Gateway.

Events reach the views shortly after a request completes. An event that arrives more than one hour after its recorded time is marked late and can change totals already shown.

Help us make these docs great!

Kong Developer docs are open source. If you find these useful and want to make them better, contribute today!