The dashboard's System area is where you see what Crew is doing to your machine and clean up after it. It has two parts: a live task-manager view of running work, and a Storage screen that reports what sessions cost on disk and lets you reclaim space.
Open System from the dashboard to see per-session resource usage as it happens — the sessions, subagents, and background work currently running, and what each is consuming. It reads like a task manager for Crew: a quick answer to "what's running right now, and what's it costing?"
Use it to spot a session or subagent that's holding memory, or to confirm the machine is idle before you start something heavy.
Crew watches overall memory pressure so heavy work fails gracefully instead of thrashing the host. When memory runs critically low, scheduled jobs defer and new subagents are refused, and the dashboard header shows the current posture so you know before you kick off something demanding.
The ceiling applies to all concurrent agents together, not per-spawn, so many small subagents can't collectively exhaust the machine. Override the aggregate limits with resource_limits.max_total_memory_mb and resource_limits.max_total_processes in config.json.
The Storage screen reports what your sessions cost on disk — transcripts, attachments, and per-session state — so a long-lived install doesn't quietly fill up. From here you can:
The session inventory reports honestly — an idle session is shown as idle rather than "in use", so the numbers you act on reflect reality.
System & storage