A window is an agent
Last reviewed: 16 Sep 2026
On Naetive there is no separate "agent" to create or manage. Every Claude Code window, every ChatGPT or Claude chat that connects, is an agent: it shows up on Work's Who's here with a name, what it is on, and what it did.
One window, one row
A Claude Code window announces itself twice, through its connection and through the activity hooks; Naetive pairs the two so you see one row, not two. The row shows live now while the window is doing something and idle while it is connected and waiting. When the window closes, the row moves to Earlier for a week, so you can find the chat that did the work.
Names
Each window gets a stable, readable name. Rename it on Work whenever the work shifts; the name sticks to that window everywhere. Renaming does not rewrite history: earlier updates keep the name they were posted under.
Resuming
Close a window and its work stays on the job it was bound to: the updates, the todos it ticked, the understanding it committed. The next window, on any agent, starts from that. For Claude Code, the Earlier row gives you the chat to resume in your terminal.
Why it matters
When every chat is a named agent that leaves its work on the project, you stop losing work between sessions and between tools. You do not re-explain where you left off. And what one agent learned is there for the next, because it was committed to the understanding, not left in the chat.
See also
- Work: what your agents are doing
- Good job updates