Analyzing and testing agents
Use built-in tools to monitor the agent's behavior and test how it responds to real user inputs before the agent goes live. These features help you find issues early, refine performance, and ensure a smooth launch.
Monitoring agent analytics
Visualize overall usage trends, track individual agent metrics like success and failed messages, and examine detailed execution paths through trace tables. For deeper troubleshooting, debug-level trace details offer granular logs to identify runtime issues. These capabilities help you quickly detect anomalies, improve response accuracy, and ensure that your agent performs reliably in production.
See more information in Monitoring agent analytics.
Evaluating a draft agent
Testing is a crucial step before going live. It allows you to simulate real user interactions and validate how your agent responds by using the testing feature. You can upload test cases that mimic user utterances, and run the evaluations to assess performance. The results provide detailed insights that help you identify gaps and improve intent recognition. With this process, you can catch issues early, refine your agent’s logic, and ensure a smooth, polished experience for users.
See more information in Evaluating a draft agent.