Usage & Cost
The Usage & Cost page helps administrators monitor LLM usage, review resource performance, configure spending limits, manage model pricing, and inspect usage events.
To access this page, open Admin and select Usage & Cost under Governance.
The page contains five tabs: Cost, Resources, Budgets, Prices, and Events.
Cost
The Cost tab displays spending across models, connections, and users for the embedded chat experience.
Filters and Actions
- Date range: Select the period for which you want to review usage. The default shown in the interface is Last 30 days.
- Connection filter: Select a specific connection or view usage across All connections.
- Reset usage: Reset usage records using the available reset action.
Usage information appears after the first chat turn has been recorded. If no usage has been recorded for the selected period and connection, the page displays an empty state.
Resources
The Resources tab displays throughput and latency information for self-hosted or local models.
Filters
- Date range: Choose the period to review. The default shown in the interface is Last 7 days.
- Connection filter: Select a specific connection or view All connections.
Resource metrics become available after the first metered chat turn. If no calls are recorded during the selected period, the page displays an empty state.
Budgets
The Budgets tab allows administrators to configure spending limits and model policies at the organization, connection, or user level. These settings are enforced server-side at the LLM checkpoint.
Enforce Budgets
Use the Enforce budgets toggle to control whether budget policies are enforced.
When enforcement is disabled, the system operates in observe-only mode. Alerts can be triggered, but requests are not blocked or downgraded by budget enforcement.
Select a Scope
Use the Scope selector to choose the level at which you want to configure a budget or model policy.
The available scope shown in the interface is Organization. Budget and model policies can be configured for an organization, connection, or user.
Configure a Budget
The Budget section includes the following settings:
| Setting | Description |
|---|---|
| Enabled | Enables the budget configuration at the selected scope. |
| Period | Select the budget period. The displayed default is Monthly. |
| Limit ($) | Enter the spending limit in dollars. |
| Alert at (%) | Specify the percentage thresholds at which alerts should be triggered. The displayed example is 80, 100. |
| On exceed | Choose the action to take when the budget is exceeded. The displayed option is Warn only. |
After configuring the settings, select Set budget at this scope to apply the budget.
Budgets use a best-effort ceiling. Requests that are already in progress when the limit is crossed are allowed to complete. Concurrent requests may cause a small overshoot, and the limit applies to subsequent calls.
Configure a Model Policy
The Model policy section controls which models are allowed at the selected scope.
- Allowed models: Configure the models permitted for the selected scope. An empty policy means all models are allowed.
- Default model: Select the model to use when a user requests a model that is not allowed by the policy.
- Set policy at this scope: Apply the model policy to the selected scope.
When policies exist at multiple scopes, the most specific scope takes precedence.
Prices
The Prices tab lets administrators maintain the per-million-token rates used to calculate usage costs.
Pricing Settings
- Currency: Select the currency used for pricing. The interface displays USD by default.
- Load template: Load a dated pricing template to populate model pricing rows for review.
The pricing table includes the following columns:
| Column | Description |
|---|---|
| Model | Model associated with the pricing entry. |
| Input /1M | Price per million input tokens. |
| Output /1M | Price per million output tokens. |
| Cache read | Price for cache-read tokens. |
| Cache write | Price for cache-write tokens. |
| Updated | Indicates when the pricing entry was updated. |
Cache-read and cache-write tokens are priced separately.
After loading a template, review the pricing rows and make any required changes. Select Save prices to save the updated rates. Use Discard to discard pending changes.
The rate card for published vendor prices is available as a separate expandable section.
Events
The Events tab provides a filtered view of recorded usage events.
Filter Usage Events
Use the available filters to narrow the displayed events:
- Date range: Select the period to review. The default shown is Last 7 days.
- Connection: Select a specific connection or All connections.
- Model: Select a specific model or All models.
- User: Select a specific user or All users.
The page also displays a summary of the number of events, total cost, and token usage for the selected filters.
When usage events exist, review them to understand recorded activity within the selected period. If no events match the selected filters, the page displays an empty state.
Notes
- Usage and resource metrics appear after metered chat activity has been recorded.
- Pricing entries affect the cost calculations displayed by AskSenseOps.
- Budget enforcement and model policies apply at the configured scope.
- Budget limits are best-effort ceilings; requests already in progress are not terminated when a limit is crossed.