Skip to main content
If you own or edit an agent that other people in your organization use, you can review its activity: who used it, how much, and — if you choose to allow it — the content of the conversations and tool calls it made.
Agent audit is an organization feature. Your administrator enables it under Settings → Organization → Audit your agent, together with two related sub-features: Audit - View conversations and Audit - View tool calls. If none of these are enabled, you only see anonymized usage numbers, never conversation content.

​
Two layers of control

Seeing what happens inside a conversation with your agent depends on two separate things being true at once:
  1. Your organization enables the audit feature. An administrator must turn on View conversations and/or View tool calls for your organization.
  2. You, as the agent owner or editor, opt in per agent. Even when your organization allows it, nothing is visible until you switch it on for that specific agent.
This means enabling audit at the organization level does not expose any conversation by itself — each agent owner decides, agent by agent, whether to share that data with editors.

​
Turning on data sharing for an agent

1

Open the agent

Open the agent you want to audit. You need Editor or Owner rights.
2

Find Editor data access

Look for the Editor data access section (“Choose which audit data agent editors can review.”).
3

Choose what to share

Turn on Share conversations with agent editors to let editors read message text. Turn on Share tool calls with agent editors to let editors see the tool calls the agent made. Each toggle only appears if your organization enabled the matching feature.

​
What you can see without opting in

Even without turning on data sharing, agent editors and owners can already see, for every conversation with the agent:
  • The conversation title and dates.
  • The number of user and agent messages.
  • The number of tool calls.
  • Total tokens used and the models involved.
This is metadata only — no message text and no tool call content. By default, you only see your own conversations with the agent; sharing is what extends this to conversations from all users of the agent.

​
Opening the audit view

Open the agent’s ⋯ menu and choose the audit action, or go to Settings → Audit your agents and select the agent from the list. Each conversation opens into a per-message breakdown showing the metadata above, plus message text and tool call details wherever your organization and the agent owner have both enabled it.

​
What end-users are told

When any sharing is turned on for an agent, WonkaChat tells the people chatting with it before they continue. They see a disclaimer badge, “Conversation auditable by agent editor,” and, depending on what is shared:
  • “This conversation is shared with the agent editors” when conversation content sharing is on.
  • “Tool calls in this conversation are shared with the agent editors” when tool call sharing is on.
This disclaimer is shown to end-users, not just to you. Only enable sharing for agents where users are expected to have their conversations reviewed — for example internal support or compliance agents — and make sure your organization’s usage policy reflects it.

​
Agent Stats

Separately from audit, Agent Stats (Settings → Agent Stats) gives agent owners and editors anonymized, aggregated usage figures for an agent: daily activity, token usage and cost broken down by model, and a transaction log — all combined across every user of the agent, without exposing who said what. Your organization must enable the corresponding feature for this tab to appear.
Your own usage across all agents is available under Settings → My Stats. Organization-wide totals are under Settings → Organization Stats, for administrators.