Agent Monitor brings observability to your ODC AI agents using a probe architecture: specialized Service Actions you drop into existing flows. Each probe captures a specific signal at a specific point in the execution chain. Nothing is recorded by default. Your team controls what gets logged and what gets redacted.

The component provides a centralized dashboard for all agents in your environment, organized by name and type. It works alongside the Agent Workbench but does not require it.

Monitor agents via Service Actions or REST APIs.

What it tracks:

• Sessions and Executions: full lifecycle of each agent run
• Model Attribution: which LLM was used at each step
• Execution Time: latency per step in the chain
• Tool Usage: Service Actions called, with inputs and outputs
• Token Consumption: input/output token counts per execution
• Input/Output Mapping: data that entered and exited each step

Developed by:

• Henrique Silva