AI Agent log detail view - Inspect a step-by-step execution log of an AI Agent run, including tool calls, inputs, outputs, and reasoning at each step.

Automation 360

Inspect a step-by-step execution log of an AI Agent run, including tool calls, inputs, outputs, and reasoning at each step.

Overview

The Agent tab in Governance logs provides a detailed, step-by-step breakdown of how an AI Agent executed. For each step in the execution, the trace surfaces the tool invoked, the inputs and outputs, timing data, execution status, and any reasoning captured from the LLM or execution engine. This level of transparency helps developers and evaluators identify inefficiencies, and diagnose failures.

Access details from the Agent Logs page by selecting View Activity Trace. Use the breadcrumb at the top of the page to return to the Governance logs landing page.

Agent runtime details


Image demonstrates the interface of Agent activities tab in Agent logs.
The Agent activities page for the agent log lists each step or tool call in execution order. Use the expand and collapse controls on each step row to show or hide its details. Expand a step to view the following details as it pertains to agent operations:
  • Step Name / Sequence Number — the name and position of the step in the execution sequence.
  • Tool Name — the tool invoked at this step (for example, Claim Assessment or API Policy Retrieval).
  • Start Time / End Time / Duration — timestamps and wall-clock duration for the step.
  • Input Variables — variable names and their values passed into the tool.
  • Output Variables — variable names and the results returned by the tool.
  • Execution Status — one of Completed, Failed, or Skipped.
  • Reasoning / Commentary — LLM-generated reasoning or execution engine notes explaining how the step result was determined, when available.

The General tab for an agent log provides a structured summary of the run's operational context and the AI model configuration used during execution. It is divided into two sections: Run Details and Model Details.

The Run Details section captures the execution metadata for the agent run. Review the following fields:

  • Status — the current state of the agent run, such as Active, Completed, or Failed.
  • Started On — the date and timestamp (with timezone) when the agent run began.
  • Ended On — the date and timestamp when the run concluded; blank if the run is still active.
  • Running Time — the elapsed wall-clock duration of the agent run.
  • Total Number of Tools — the count of distinct tools registered or invoked during the run.
  • User — the user account under which the agent run was initiated.
  • Activity Id — a unique identifier for the agent run, useful for cross-referencing logs or raising support requests.

Image demonstrates the interface for General tab in Agent logs.

The Model Details section captures the AI model configuration and token consumption associated with the run. Review the following fields:

  • Publisher — the organization that provides the underlying AI model (for example, OpenAI).
  • Model — the specific model version used during the run (for example, GPT-4.1).
  • Model Connection Type — the connection method category, such as Standard.
  • Model Connection Method — the authentication approach used to connect to the model, such as API key.
  • Model Connection Name — the configured connection profile name as defined in the platform.
  • Tokens Used — the total number of tokens consumed during the run.
  • Request Configuration — the model parameters applied to the request, including settings such as temperature, max_tokens, and tool_choice.
  • Response Attributes — a breakdown of token consumption returned in the model response, including prompt tokens, completion tokens, and total tokens.

Filtering

Use the filter controls at the top of the Agent Logs to narrow the steps displayed:
  • Status – filter by Completed or Failed execution status.
  • Activity Name – filter by the specific execution.
  • Agent Name – filter results based on the name of the individual agent.
  • Source – filter agents based on the source of the agent.

Permissions

The Agent Log view is available to users assigned either of the following permissions:
  • View AI Governance Logs
  • View AI Governance Log Details