Agent Builder → Audit Log records every action taken by every assistant. It is the primary tool for debugging problems and reviewing what your AI assistants have actually done.
The Stats Bar
At the top of the page a summary bar shows totals for the last 30 days:
- Actions — total number of recorded events
- Tokens Used — cumulative LLM tokens consumed across all assistants
- Estimated Cost — approximate spend at your provider’s standard pricing
Filtering
Two dropdowns let you narrow the view:
- Agent — show entries for one specific assistant only
- Action — filter by event type (for example
task_start,task_complete,task_error,chat_message)
The page displays the most recent 100 log entries after filtering is applied.
What Gets Logged
The audit log records all significant assistant activity, including:
- Chat messages sent to and received from the AI provider
- Scheduled task starts, completions, durations, and errors
- Event listener triggers and their outcomes
- Tool calls made during a session (file reads, database queries, code proposals)
- License validation checks
Each entry includes the assistant ID, action type, timestamp, token count, and any relevant metadata (such as error messages or the number of tool calls in a session).
Log Retention
Old log entries are cleaned up automatically by a daily scheduled task. The default retention period keeps entries for 90 days. You can adjust this in the plugin’s database configuration if you need a longer or shorter history.
Debugging with the Audit Log
If an assistant behaves unexpectedly, filter the log by that assistant and look for task_error or event_error entries. The detail metadata will usually contain the error message returned by the AI provider or the PHP exception that caused the failure.