Early stage · Go

boxr

Linux containers, down to the kernel.

A small, daemonless container runtime built from scratch in Go with Linux namespaces, cgroup v2 and OverlayFS. It exists to make the kernel primitives behind containers understandable, while still giving useful workflows.

What it does

Namespaces

PID, UTS, mount, IPC and user namespaces, with host, isolated or managed bridge networking.

cgroup v2

Memory, CPU, I/O and PID limits applied before the workload runs.

Images

Local OCI image store with pull, push, build, verify, save and load.

Volumes

Named volumes with labels, export, import and safe pruning.

Deployments

Dependency-aware multi-service manifests and a Compose subset.

Hardening

Capability drops, no-new-privileges and a seccomp baseline.

Try it

sudo boxr doctor
boxr run --dry-run --name preview -- /bin/echo hello
sudo boxr run -d --name worker -- /usr/bin/sleep 3600
boxr ps
boxr logs --follow worker

Not a hardened security boundary or a Docker replacement. Don’t run untrusted workloads with it.

Live repo stats

starsissuesrelease

Built with