I got tired of hunting through AI coding sessions, so I put them on a physical control deck
I’ve been running multiple coding agents and sessions at the same time, and one problem kept getting more annoying:
figuring out which session was working, which one was done, and then finding the right terminal again.
So I’ve been building a physical control surface for that workflow.
This short demo shows the part I’ve been working on recently:
- The LCD shows 3 sessions for the currently active coding agent
- K1–K3 jump directly to those sessions
- The top row switches between active coding agents
- When the active agent changes, the LCD session list and available actions change with it
- The LCD footer shows the current context-sensitive shortcut layout
The goal isn’t really to build another macro pad.
What I’m experimenting with is moving agent state + session state + common actions out of terminal tabs and onto a persistent physical interface.
I’m currently testing workflows with Qwen, DSH, ChatGPT, Herdr and OpenCode.
The biggest thing I’m still thinking about is how much of the control surface should be fixed vs context-aware.
If you regularly run multiple coding agents, would you rather have fixed shortcut keys, or keys that automatically change with the active agent/session?