Concepts
Usage
Monitor every run with detailed logs, token metrics, and action traces.
Overview
Usage is your operational visibility layer. It tracks assistant runs, token usage, response time, and action execution details so you can understand exactly what happened in production.
When something fails, you can see where and why. When behavior is expensive, you can see which runs and models are driving cost.
What you can inspect
- Per-run logs with model, input tokens, output tokens, and latency.
- Message-level traces for each interaction.
- Action logs including parameters, results, success/failure, and duration.
- Filters by assistant for faster debugging.
Why this saves time
- Faster root-cause analysis when answers are wrong or incomplete.
- Less guesswork during prompt or tool changes.
- Better cost control with token-level visibility.
- Easier handoff between product and engineering teams.
Build links
Use case slots
Use case slot 1
Add your real-world monitoring use case here.
Use case slot 2
Add your real-world monitoring use case here.