We designed Paygent around something we call indicators. It is the main building block. Lightweight, event-driven, and built for how agents actually run.
What's an indicator?
An indicator is a named metric your agents emit. Each event has a value, such as tokens_used or call_duration, plus optional metadata like model or customer_id. Paygent aggregates these as they come in for cost tracking and billing.
Why we went this route
- Flexible: Define any metric that matters. Tokens, API calls, outcomes. Your call.
- Real-time: No batch jobs. Aggregation happens as events arrive.
- Multi-tenant: Attribution by customer, agent, or whatever dimension you need. Handles that automatically.
Integration
Drop a few lines into your agent pipeline and you're emitting. No heavy middleware, no polling. We tried to keep it as lean as possible.
