Skip to content

AI Usage History

AI Usage History is a read-only log of every AI API call made within the tenant. Tenant Super Users use it to investigate costs, identify high-token requests, review usage by user or service, and verify that AI budget controls are working as expected. Each row represents one AI API call and records the provider, model, token consumption, and both the raw provider cost and the amount charged to the tenant.

AI Usage History page showing a table of AI call records with columns for date, user, service, provider, model, input tokens, output tokens, cost, and charged amount. The demo tenant shows "No usage records found."

  • Route: /ai-usage
  • Menu Path: Settings → AI → AI Usage History
  • Primary audience: Tenant Super Users and tenant administrators.
  • Browse the full history of AI API calls made by all users and automated processes in the tenant.
  • See each call’s provider, model, input token count, output token count, provider cost, and charged amount.
  • Identify calls made by system processes (shown with a System Call badge) rather than named users.
  • Paginate through the history using traditional page controls or adjust the page size.
  • Compare provider costs against charged amounts to understand margin or markup applied via AI model pricing settings.
  1. Open Settings → AI → AI Usage History.
  2. Review the table to identify the most recent AI calls.
  3. Look at the Charged column to see the total cost impact per call.
  4. If a specific call looks unexpectedly expensive, note the Model and Input / Output Tokens columns to understand why.
  5. Sort or paginate through the list to find calls from a specific time window.
  6. Cross-reference high-cost models against the AI Model Pricing page to confirm per-token rates.
  • AI Usage History records are immutable. There is no create, edit, delete, or export action on this page.
  • An empty list means the tenant has made no AI API calls. The demo environment shows “No usage records found” because no AI calls have been initiated against it.
  • The list uses traditional page-based pagination (not Load More). You can choose page sizes of 10, 20, or 50 rows.
  • The total count of records matching the current page state is shown above the table.
  • System-originated calls (triggered by background jobs or automation rather than a named user) display a blue System Call badge and a server icon in the User column instead of a user name.
  • Costs are displayed in USD with four decimal places of precision (e.g., $0.0042).
  • Provider Cost reflects what the AI provider charges. Charged reflects the amount applied to the tenant billing, which may differ if markup or custom pricing is configured in AI Model Pricing.
ColumnSortableDescription
DateNoTimestamp of the AI call in MMM D, HH:mm format
UserNoName of the user who triggered the call; system-originated calls show a server icon and “System”
ServiceNoName of the portal service or module that made the call (e.g., core, a registered microservice name)
ProviderNoAI provider name (e.g., OpenAI, Anthropic)
ModelNoModel identifier in monospace font (e.g., gpt-4o, claude-3-5-sonnet)
Input TokensNoNumber of input (prompt) tokens consumed, formatted with thousands separators
Output TokensNoNumber of output (completion) tokens consumed, formatted with thousands separators
CostNoRaw provider cost in USD with four decimal places
ChargedNoAmount charged to the tenant in USD with four decimal places
System CallNoBlue badge with server icon shown when the call was made by an automated system process rather than a named user

The page uses traditional pagination with configurable page size. Available page sizes are 10, 20, and 50 rows. The total count of records is displayed above the table. Navigate using the Previous / Next controls or jump to a specific page number.