Agent Execution Audit Trail

Introduction

The Agent Execution Audit Trail is used to view the audit trail of all agent executions with detailed trace information. This feature provides complete visibility into all agent executions, offering detailed trace information to help developers, administrators, and auditors monitor, troubleshoot, and optimize agent behavior.

Agent Execution Audit Trail

The Agent Execution Audit Trail serves as a centralized record of every agent run. It captures execution details in a structured format, enabling teams to:

• Track activity: Understand when and how agents were executed.
• Diagnose issues: Pinpoint errors or performance bottlenecks with precise trace data.
• Ensure compliance: Maintain accountability with user and process-level information.
• Optimize workflows: Analyze execution metrics to improve efficiency and reliability.

Information available in the Agent Execution Audit Trail

• Start Time: Timestamp marking when the execution began.
• Agent Name: Name of the executed AI agent.
• Agent Type: Either AI Agent Builder or AI Composer.
• App ID: Identifier of the application the agent belongs to.
• Run ID: Unique identifier for the execution instance.
• Status: Final outcome of the execution. Either RUNNING / SUCCESS / FAILED / ABORTED.
• Username: The username of the individual who triggered the agent.
• Full Name: The full name of the user associated with the execution.
• Node Count: Number of nodes traversed during the execution.
• Input Tokens: Prompt tokens used.
• Output Tokens: Completion tokens used.
• Duration: Total time taken for the execution to complete.

Agent Execution Detail

Each entry in the audit trail provides comprehensive metadata about the agent run. This ensures that every execution can be traced back to its origin and analyzed in detail.

• Run ID: Unique identifier for the execution instance.
• App ID: Identifier of the application the agent belongs to.
• Start Time: Timestamp marking when the execution began.
• Duration: Total time taken for the execution to complete.
• Node Count: Number of nodes traversed during the execution.
• Username: The username of the individual who triggered the agent.
• Full Name: The full name of the user associated with the execution.
• Process ID: Identifier of the process under which the agent was executed.
• Status: Final outcome of the execution. Either RUNNING / SUCCESS / FAILED / ABORTED.