Monitor is where you look after a running thunk. Its tabs let you stop and restart the thunk, send events about its work to your own tools, save reports, and see its errors, AI spend and timing. Open it once a thunk is doing real work.
Open Monitor
Open your thunk.
In the left navigation, under Admin, select Monitor.
Select a tab across the top of the pane.
Monitor is available to the thunk's owner and admins. In a thunk that only holds content folders, Monitor is greyed out, because the thunk runs no work. For more on thunk roles, see Users and access controls.
The tabs
Tab | What it is for | Read more |
Pause / Restart | Shows whether the thunk is Live or Paused, and stops or restarts all of its AI work at once | |
Observability | Notifies your own monitoring or analytics platform each time a work item finishes | |
Reporting | Lists the reports saved from filtered views of your work items, and the report automation that sends them on a schedule | |
Errors | Errors people have run into while the thunk was running: a daily error-rate chart, then the most recent errors, newest first | |
Token Costs | What the thunk spends on AI each day, for the whole workflow or for one step | |
Time Analysis | How long the thunk's work takes, for the whole workflow or for one step, and which work items were slowest | |
Automation Log | A log of the work that happens behind the scenes to keep the thunk running | See below |
Automation Log
Most of a thunk's work happens inside the steps of its work items, and you read that work by opening a work item. Some work belongs to the thunk as a whole rather than to one work item, and the Automation Log records that work.
Check it when something the thunk should do on its own did not happen and no work item shows why. For example, when an observability event cannot be delivered, the log records why, such as the HTTP status that came back. See Observability and Monitoring.
Where to start
Monitor shows each part of a thunk's health on its own tab. It does not tell you which tab to open first. When you are checking how a running thunk is doing, start with stuck work and errors. Look at time and cost only once the thunk is doing the right thing. Checking a thunk's health: where to start walks through the checks in order.