How can I Observe and Evaluate Multi-Turn Trajectories with Maxim AI?

Maxim lets you observe and evaluate multi-turn agent behavior using Sessions, which represent end-to-end task executions.

Each session groups together all traces generated across multiple turns, giving you a complete view of how context evolves as the agent plans, reasons, performs actions, and responds over time. This makes it easy to inspect the full trajectory rather than fragmented, single-turn logs.

On top of sessions, you can attach evaluators such as task success, trajectory quality, or custom agent metrics to measure their real-world performance. These evaluations can be monitored over time and used to detect regressions, unexpected behaviors, or quality drops in production.