every 5dive agent already has a live terminal. the problem was seeing them all at once.

v0.46 adds 5dive wall.

5dive wall
5dive wall --grid=3x2
5dive wall main quinn olivia ops dev --grid=3x2

every running seat, tiled on one screen.

open that full size if you want to read what the seats were doing.

plain 5dive wall picks up the running tui-enabled seats on the box. name them if you want a specific wall in a specific order, add --grid= if you want a specific shape.

the sixth tile up there is empty on purpose. only five seats were running, and the wall keeps the slot instead of reflowing five tiles into a layout you have to re-learn.

spare slots stay open. when another seat comes online, you can fill one without restarting the wall:

tmux respawn-pane -k -t wall.<pane> "5dive wall --follow <seat>"

--follow keeps that tile attached to the seat, including through restarts.

the wall starts read-only, and that is a safety property rather than a preference. a Ctrl-C into an agent’s pane kills that seat’s unit. C-b w opts ONE pane in, C-b r locks it again, C-b z zooms it, C-b d leaves. the agents keep running.

one thing to know before you run it: each seat’s terminal belongs to that seat’s own user, so the wall reaches it as that user or not at all. run it as root, or grant the attach on its own. without that, every pane tells you NOT PERMITTED rather than quietly drawing you a dead company.

the useful bit is that these aren’t copies or log streams. wall attaches to the tmux sessions the agents are already working in.

you could always open any seat directly:

5dive agent <name> tui

useful for one agent. less useful when the whole team is working at once.

that’s the whole feature: the team was already running. now you can look at all of it.

---

5dive runs agents on your own server, on subscriptions you already pay for. start here, or read the code: github.com/5dive-ai/5dive.