Blog
Run Claude Code Across Every Machine You Own
If you have more than one Mac, you've felt this: you start a Claude Code session on the laptop in the morning. By afternoon you wonder if you should've started it on the GPU rig. By evening you're SSHing into all of them in turns.
Celistra was built for this exact problem. Read the full multi-machine setup guide. The summary:
- One
celistradbinary on every machine you own - 30-second pairing per machine via tray icon
- One dashboard with a unified fleet view
- Spawn Claude Code (or Codex, Aider, anything) on whichever box has the right context
- Watch from your phone
- Approve permissions with haptic taps
- Sub-millisecond on LAN; transparent tunnel fallback off-LAN
- Sandboxed by default — workspace-only writes, capability tokens for everything else
The competing options:
- Claude Code Remote Control — official Anthropic feature, single session per machine, terminal must stay open. Detailed comparison.
- SSH + tmux — works for one or two boxes; falls apart at three. Detailed comparison.
- K3s / Nomad — way too heavy for this use case. Why Kubernetes is overkill.
The flow
- Download daemon: celistra.dev/download
- Install on each machine; tray icon appears
- Open dashboard, sign in with Google
- On each machine, click tray → Pair this machine
- Done. Spawn agents from anywhere.
Free for 1 node
Pro is $2/month for 5 nodes. Team is $10/month for unlimited nodes plus RBAC. The daemon is free regardless.
FAQ
Will it work on my Mac mini under the TV?
Yes. Mac, Linux (incl. Pi), Windows.
Does it handle GPUs?
It doesn't manage GPUs — it spawns processes that can use them. If your training job uses CUDA, Celistra spawns it; CUDA still works the same.