Usage

The Usage tab shows how an agent is doing for you over the last 30 days. Pick an Agent at the top; the refresh button reloads the numbers.

Tiles

Tile	What it counts
Runs	Your chat responses and tasks in the period, split into Succeeded, Failed and Other outcomes.
Completion rate	Succeeded out of finished runs; excludes active runs.
Tool attempts	Tool calls the agent attempted across the runs shown.
Average duration	From the finished runs shown; excludes queue time.

Where your work goes

A flow from Work (the kind of run, such as chat responses or tasks) through the agent to each Outcome. Ribbon width represents runs, not tokens or tool calls. Select an outcome to filter the run history below.

Run history

Every run in the period, newest first, with Conversation or task, Outcome, Started, Tool attempts, Duration and Open run to see the full run.

tip

Agent tool calls go through your MCP Gateway like any other tool call, so they also appear in the MCP Gateway activity for that server.