Skip to main content

What is a Trace?

A trace is a single AI interaction captured within a session. Every message sent to an AI service, every response received, and every event triggered creates a trace. Traces are the raw data that agents evaluate. Each trace captures what happened, where it came from, and what the outcome was.

Trace Directions

Every trace has a direction that describes the type of interaction:
DirectionDescriptionExample
InletAn incoming request to an AI serviceA user submits a prompt to ChatGPT
ResponseAn outgoing response from an AI serviceThe chatbot’s reply
SignalA passive event or notificationA tool call, system log, or background process

Viewing Traces

Within a session, click any trace to see its full details:
  • Direction and source indicate the type of interaction and which service was involved
  • Agent evaluations show what each agent found
  • Status and outcome show whether the trace was allowed, blocked, or escalated
  • Related traces provide other traces in the same session for context

What Happens After a Trace Is Captured

1

Agents Evaluate

Your active agents evaluate the trace against your compliance requirements. Each agent assesses the trace independently.
2

Action Resolved

Based on the agents’ findings and their configured roles, the platform determines the outcome: allow, block, or escalate.
3

Status Updated

The trace is marked as Completed, Rejected, or Escalated. If escalated, a review task is created for your team.
For the full details on how this works, see Trace Lifecycle.

Sessions

Understand how traces are grouped into sessions

Trace Lifecycle

Follow a trace through the evaluation process