Running tasks
While a reply is running, a live line shows what the agent is doing, such as Thinking, Running a command or Working. Under it, a caption shows elapsed time, the reported token count and any running tasks. For example: 52s · 1.2k tokens · 1 running task.
The token count comes from the engine's own usage report. Crew does not estimate it from the length of the reply. Large counts use short forms such as 1.2k; until the engine reports a count, the caption shows — tokens.
The task list above the message field
Background tasks and subagents appear in a small list pinned above the message field, for example 1 task running or 2 running · 1 done.
- Click the line to expand it and see each task, with the time since Crew first saw it. Up to four rows show, then and N more.
- Move the pointer over a running task and choose Stop to stop it. With two or more running, Stop All asks once, then stops them all.
The list hides 10 seconds after the last task ends.
Read a command's output
Open the conversation's ⋯ menu and choose Files & Terminals…. Under Terminals, a running background command, or one that left output, has Output (and Hide) to read its latest output, which you can select and copy but not type into. A running command also has Stop…, then Stop it or Keep. Choose Done to close. The item appears only when the conversation supports it.
A task row also shows View Live Output on hover, but in this version it does not open a view yet; use Files & Terminals… instead.
While Presenting is on, private text in task names and output is hidden.
When the reply ends
Background work can continue after the reply finishes. The completed turn then says that a task is still running. A task row's elapsed time starts when Crew first sees that task, so it may be shorter than the task's actual running time.