Back

01 / 05

Agent Deck

A macOS workbench for directing many Claude Code agents at once: see at a glance what each session is doing and whether it’s waiting on me.

  • Electron
  • React
  • TypeScript
  • Claude Agent SDK
  • Monaco
  • xterm.js
agent-deck — sessions
The Agent Deck window: sessions listed on the left; on the right, a session waiting for approval to run a command

What I built

One window shows what every Claude Code session is running, where it’s stuck and whether it needs my approval. Anything I have to decide (approving a command, picking an option) lands in one place, so I don’t hunt through windows.

Sessions can run on other machines, such as a GPU server running training, and I can still watch and drive them from this Mac.

A built-in plan library keeps the tasks, the decisions made and the hand-overs, so a fresh session can read it and carry on.