Once you have more than 1 agent running, its really hard to answer the simple question: what is it doing right now and why?
It is a local terminal UI that collects your agents runtime events and shows their activity in one place.
You can inspect:
- agents and subagents
- tool calls
- prompts
- outputs
- status changes
Sessions and projects are there to give that activity some context, so it is easier to follow what happened over time and which run belongs to which repo. And make it easy to see all the agents involved in a run at a glance and understand what each one is doing.
If you already use Claude, Codex, or OpenCode heavily, I would love to know if this feels useful and what you would want to see.