Skip to main content
Observability is critical for understanding what your agents and tools are doing. It gives you visibility into performance, reliability, and behavior across your MeshAgent rooms, so you can debug faster, track performance, and build trustworthy and transparent agentic applications. As agents take on more complex and autonomous tasks, it’s increasingly important to know exactly what they’re doing so you can see where tokens are being spent, how much time tasks take, how effectively agents use tools, and which tools, participants, or rooms are most active.

​
Why Observability Matters

Observability is essential at every stage — whether you’re building, testing, or scaling your agents. When you’re developing, it helps you confirm that everything is working as expected and understand how your systems behave. As you scale, those same insights help you detect issues early, measure performance, and make confident improvements. No matter how you’re using agents, you need detailed, real-time visibility into what’s happening so you can understand what’s going well, what’s slowing down, and where things might be breaking. MeshAgent captures this information through telemetry — structured data about what your agents, tools, and rooms are doing as they run. Under the hood, MeshAgent uses OpenTelemetry, an open standard for collecting and analyzing telemetry data such as:
  • Traces — the complete path a request takes through your agents and tools
  • Spans — the individual steps within that path (like TaskRunner.ask or Tool.execute)
  • Metrics — performance and usage counters (e.g. response times, token and cost tracking)
  • Logs — contextual messages that help you understand what happened and why
Together, these details give you a full picture of how your system behaves so you can fix issues faster, improve reliability, and scale with confidence. MeshAgent provides rich telemetry out of the box that you can extend with your own custom traces, logs, and metrics to monitor what matters most in your workflows.

​
What MeshAgent Provides

MeshAgent includes built-in observability so you get detailed, structured insights by default. At the project level, the Usage tab in MeshAgent Studio shows aggregated insights like:
  • Cost by provider
  • Most active rooms and participants
  • Most used tools
  • Tool call latency
At the room level, each session automatically records:
  • Logs for room, agent, and tool activity
  • Traces to understand task execution and tool calls
  • Metrics for insights into session costs
You can explore these live in the room’s Developer Console or it’s Session Viewer.

​
Next Steps