Skip to main content
The Usage Breakdown view offers detailed insight into how Kubiya’s agentic workflows consume resources over time. It highlights total conversation turns, runtime duration, cost, and per-turn efficiency, enabling teams to monitor consumption patterns and evaluate operational efficiency at both macro and micro levels. Kubiya Platform Overview This page complements Agentic Engineering Minutes (AEM) by translating AEM-based computations into tangible runtime and cost data. It provides a clear picture of how agent executions translate into cost-per-turn and model-tier utilization.

​
Key Metrics

1. Total Turns: Represents the total number of conversation or execution turns completed within the selected time period.
  • Value: 12 turns
  • Each turn corresponds to one complete interaction or execution cycle within a task or workflow.
2. Runtime: The cumulative duration of all runs, measured in minutes.
  • Value: 20.1 total minutes
  • Average: 1.67 minutes per turn This metric reflects the actual operational time agents spent processing workloads.
3. Total Cost: The aggregate AEM cost associated with all runs during the selected timeframe.
  • Value: $3.01 total
  • Calculated using the AEM pricing rate of $0.15 per AEM minute. This value helps quantify the financial footprint of agent operations.
4. Average Cost per Turn: Represents how much each turn costs on average.
  • Value: $0.25 per turn Useful for assessing cost efficiency and projecting future workload expenses.
The Usage Trends graph visualizes the correlation between runtime, total turns, and cost over time. It allows users to identify growth patterns, peaks in usage, or consistent consumption behaviors.
  • Runtime (min): Tracks how long agents are running per day or week, revealing efficiency and throughput.
  • Turns: Reflects activity volume and engagement frequency across agents.
  • Cost ($): Maps operational expenses proportionally to runtime and turns.
For this sample period, trends indicate steady growth in usage across both runtime and cost, confirming consistent system utilization without performance anomalies.

​
Model Usage Distribution

The Model Usage Distribution section breaks down consumption by model capability tier, helping teams understand which model types dominate workloads and costs. In this case:
  • Model Tier: Mid (Claude Sonnet 4)
  • Turn Distribution: 100%
  • Cost Distribution: 100%
  • Runtime: 20.1 minutes
  • AEM Consumption: 20.1 AEM
  • Cost Efficiency: $0.25 per turn
This indicates that all tasks were executed exclusively on the mid-tier model (Sonnet), showing uniform workload assignment and predictable cost-per-turn behavior.

​
Interpretation and Optimization

The Usage Breakdown data provides a practical basis for optimizing both runtime and cost efficiency:
  • Monitor per-turn runtime: High per-turn averages may indicate complex or inefficient workflows; breaking them into smaller, modular processes can improve throughput.
  • Review cost patterns: Align frequent or simple operations with lower-weight models (e.g., Haiku) to reduce AEM costs.
  • Balance task load across tiers: Distributing workloads by complexity ensures optimal utilization of model tiers while keeping expenses predictable.
  • Track usage trends: Regularly reviewing trend data helps anticipate spikes in operational demand and manage budgets proactively.