Audit activity of an AI Agent and interactions between its model connection to support governance, security, and responsible AI compliance with AI Agent Governance Logs.
Overview
AI Agent Governance Logs provide a complete, step-by-step audit record of every interaction between an AI Agent and its model connection (model) throughout an agent execution. Each log captures inputs and outputs at every step, including tool call requests and responses, initial system prompts, variable definitions, and token usage.
Logs are generated from the moment an AI Agent execution begins and remain accessible while the execution is in progress and after it terminates. Because an agent execution can span days or weeks, governance admins can review the accumulated log at any point during or after the run. Logs are retained for 180 days.
How It Works
As an AI Agent runs, each new interaction with its model is recorded as a discrete step.
To access the Agents tab, navigate to , then select Agents. The table displays the following columns:
| Column | Description |
|---|---|
| Status | The current state of the agent run. Values are: Active, Completed, Cancelled, or Failed. |
| Activity name | The name of the AI Agent task followed by a reference ID, in the format Agent task name - Ref:X-Y. This matches the activity name shown in the activity reporting UI. Click the activity name link to open the detail view for that execution. |
| Agent name | The file name of the AI Agent as stored in the repository. |
| Source | The origin of the agent execution. For executions triggered directly through the AI Agent, this displays AI Agent. |
| Running time | The elapsed time from the start of execution to termination (Completed, Cancelled, or Failed). For runs still in progress, this reflects the time from start to when the page loaded. |
| Started on | The date and time the agent run began, displayed in the viewing user's configured timezone. |
| Initiated by | The name of the user who started the agent execution. |
Select View Activity for any listed agent execution to open the detail view. This lets you drill down into each step of a specific execution.
- System prompt and user prompt
- Tool definitions, tool call requests, and responses
- Model configuration parameters and token counts
- Session metadata: session ID, device name, run-as user, and start and end timestamps
MCP Inbound tool calls
- Tool name
- Input and output parameters
- Session ID
- Device name
- Run-as user
Benefits
Governance logs give compliance and oversight teams the visibility they need to verify that AI Agents are operating within defined boundaries. By exposing every LLM interaction in a structured, navigable format, the logs reduce the time required to investigate anomalies, prepare compliance reports, or respond to audit requests.
Because logs are available during execution, administrators can identify issues in long-running agents before they escalate. The 180-day retention period supports both routine auditing and retroactive investigation.
Requirements
The following requirements must be met before you can use AI Agent Governance Logs:
- License: An Enterprise Platform license is required. For information about supported versions for this feature, see Enterprise Platform.
- Permission: Users must be assigned the View AI Governance Log Details permission. An Automation Admin assigns this permission via .
- Data logging: Data logging must be turned on in the Control Room.
- AI Agent deployment: At least one AI Agent must be configured and have completed at least one execution before logs are available to view.
- Retention period: Logs are retained for 180 days from the date of the agent execution. Executions older than 180 days are no longer accessible in the AI Governance log view.