Why cost controls matter#
AI actions consume tokens, and token costs can add up quickly — especially with multi-step workflows and large context windows. Cost controls prevent surprise bills.
Budget levels#
Workspace budget#
Set a monthly spending cap for the entire workspace:
| Setting | Description |
|---|---|
| Monthly budget | Maximum spend per calendar month |
| Alert threshold | Percentage at which you're notified (default: 80%) |
| Hard stop | Whether to block actions when budget is exceeded |
Per-action limits#
Cap individual action costs:
| Setting | Description |
|---|---|
| Max tokens per action | Prevents single expensive calls |
| Max cost per action | Dollar-denominated limit |
| Daily action limit | Maximum actions per 24-hour period |
How limits are enforced#
- Before executing an action, LumenFlow estimates the token cost
- If the estimate exceeds limits, the action is blocked
- The user is notified with the estimated cost and the limit
Estimation uses the input token count and a model-specific multiplier for output tokens.
Monitoring spend#
Track costs in Observe → Analytics:
- Real-time spend for the current billing period
- Breakdown by action type and model
- Daily and weekly trends
- Forecast for the remainder of the month
warning Cost estimates are approximations. Actual costs depend on the provider's pricing and the model's actual output length. Set budgets with a safety margin.