Docker Sandboxes on Raspberry Pi
Run AI coding agents inside real microVMs on a Raspberry Pi 5. The Pi's Arm cores support hardware virtualization through KVM, so sbx gives you the same isolation boundary here as on a laptop.
Every output marked Real output in this track was captured on the hardware below.
| Component | Used for this track |
|---|---|
| Board | Raspberry Pi 5 Model B Rev 1.1, 8 GB RAM, 4 cores |
| OS | Ubuntu 26.10 (arm64), kernel 7.3.0-1003-raspi |
| Storage | 64 GB-class microSD card (58 GB usable) |
| Cooling | CPU cooler with fan |
| Docker Sandboxes | sbx v0.54.0 |
Not in the main lab's platform table
The sbx pre-flight checklist lists Linux x86_64 but not Linux arm64. This track covers arm64 as tested on one Pi 5. Other Pi models and operating systems were not tested.
You don't need Docker Engine
Docker Engine was not installed on the Pi used here. Each sandbox ships with its own Docker daemon inside the microVM.
What you'll do
- Set up the Pi - check KVM, install
sbx, sign in - Create your first sandbox - boot a microVM
- Prove the isolation - kernel, files, and Pi hardware
- Control the network - allow and deny hosts live
- Run Docker inside the sandbox - containers and ports
- Fit it on 8 GB - memory, heat, and parallel sandboxes
- Run Claude Code and clean up