Token Usage
Token control in Devic allows you to analyze the real resource consumption during agent execution.Each interaction with the model—whether it’s data input, response generation, or tool usage—is recorded to provide an accurate view of the operational cost and agent efficiency.
Consumption Breakdown
Hovering over the Tokens field of any execution displays a summary with detailed usage:| Metric | Description |
|---|---|
| Input Tokens | Tokens consumed by input information (prompts, context, knowledge bases, etc.). |
| Output Tokens | Tokens generated by the model as a result or final response. |
| Total Tokens | Total sum of input and output tokens. |
| Cached Tokens | Tokens reused through the caching system, which are not recomputed in active consumption. |
Cache System
Devic implements a token caching system that automatically optimizes resource usage.When an agent reuses information or fragments already processed (for example, base instructions or RAG context), tokens are retrieved from the cache without generating additional cost. This allows:
- Reducing total token consumption.
- Speeding up responses.
- Maintaining contextual coherence in consecutive executions.
Consumption Analysis and Optimization
The execution panel provides an overview of the agent’s economic and technical performance.These metrics allow developers and operations teams to:
- Identify processes or tasks with higher average cost.
- Adjust prompts, configurations, and tools to improve efficiency.
- Compare consumption between different versions or agents.
- Measure savings achieved thanks to caching.
Detailed token usage tracking in Devic provides complete traceability of operational cost, enabling informed decisions about efficiency and scalability in production environments.
Next step
Explore advanced configurations such as subagents, scheduled executions, or contextual control strategies for your agents.