Agentflow execution history

The Agentflow execution history page provides observability into agentflow executions. Use it to review how an agentflow action was processed. The execution history includes the input, output, tools, execution time, and token usage for each action. You can use this information to understand how the agentflow processed data, investigate unexpected results, review the data passed between actions, and troubleshoot an agentflow execution.

View the agent execution history

You can view analytics for all agentflows from the Instances page.

1. Open the agentflow Instance details. For more information, see Access the Instance details page.

2. In Instance activities, expand the agentflow action.

3. Click Open agentflow execution history.
   The agent execution history opens in a new browser tab.

The left pane displays the actions included in the execution, such as Start and Agent. Select an action to view its execution details. A status icon indicates whether the action completed successfully.

View the Start action

The Start action displays the data used to initiate the agentflow execution. You can view the action data in the following formats:

• Rendered: Shows the action input and output in a readable format. The Rendered view contains the following fields:

  • Input: Data provided when the agentflow execution started. If no input was provided, the field displays None.

  • Output: Data produced by the Start action and passed to the next action. If the action did not produce output, the field displays No data.

• Raw: Shows the complete action data as a structured object. Expand an object or field to view its contents.

View the Agent action

On the agent execution history page, click Agent in the left pane to review the agent action details. The elapsed execution time and token usage appear in the upper-right corner of the panel.

You can view the agent action data in the following formats:

• Rendered: Organizes the execution information into readable sections. The Rendered view can contain the following sections:

  • Tools: Lists the tools available to or used by the agent during the execution. Expand a tool to view additional information when available.

  • Input: Shows the information provided to the agent. The input can include system instructions, user input, and information generated while the agent processed the request.

  • Output: Shows the result produced by the agent action.

  • Tokens: Shows token information for the agent action, including the configured token limit and token limit status.

    Note: The information displayed depends on the agent configuration and the execution. Some fields can be empty or unavailable.

  • Token limit: Maximum token limit configured for the agent action.

    Note: Token utilization behavior might differ depending on the model.

    • Anthropic models might report token utilization greater than 1.0 when generated tokens exceed the configured token limit.

    • Gemini models might end response generation because the token limit is reached even when token utilization is slightly less than 1.0.

    • Token limit status: Indicates the current token limit status. Possible values are Within limit, Approaching limit, and Limit reached.

• Raw : Shows the complete agent action data as a structured object. Expand an object or field to view its contents.

Agent executions page fields and settings

The Agent executions page displays agentflow executions in the environment. Use this page to filter executions by state, date, or agentflow identifier and review execution status, session details, and timestamps.

Selection	Description
State	An execution state to filter the list. All: Shows executions in all states. In Progress: Shows executions that are currently running. Finished: Shows executions that completed. Error: Shows executions that encountered an error. Terminated: Shows executions that were terminated before completion. Timeout: Shows executions that exceeded the allowed execution time. Stopped: Shows executions that were stopped before completion.
Start date	The earliest creation date to include in the results
End date	The latest creation date to include in the results.
Agentflow identifier	An agentflow identifier to display executions for a specific agentflow. Use the X icon to remove the filter.
Apply	Apply the specified filters to the list.
Reset	Clear all filters and display the default list of executions.
Status	Current state of the execution, such as Complete.
Last Updated	Date and time when the execution was most recently updated.
Session	Unique identifier for the session associated with the execution.
Created	Date and time when the execution was created.
Actions	Click to view the available options. View Details: Opens the agent execution panel, where you can review the execution details.