Namespace is the one thing kush lets you change without leaving your shell. Context stays locked for the life of a pinned shell, that's the whole design, but you'll constantly want to move between namespaces in the same cluster, checking billing and then kube-system without re-entering anything. kush ns is how.

Re-pin the namespace in place

Inside a kush shell, kush ns <name> rewrites the namespace in that shell's private kubeconfig where it sits. No new shell, nothing lost from your session, and it applies on the very next command:

Re-pin a running kush shell to billing

Because it edits the existing kubeconfig instead of spawning a shell, kush ns is exempt from the nesting guard. It's the one kush command you can run from inside a kush shell.

From a normal shell, it opens one for you

Run kush ns <name> when you're not already pinned and it spawns a fresh shell for your current ambient context, set to that namespace. Read it as "give me an isolated shell on the cluster I'm already pointed at, in this namespace":

Namespace subshell from outside

$ kubectl config current-context
staging

$ kush ns billing
# now in a subshell pinned to staging/billing

No name? kush asks

Leave the name off and kush prompts for one as free text:

Prompt for a namespace

$ kush ns
namespace: billing

It's a plain prompt, not a picker, and that's on purpose: listing namespaces would mean a call to the cluster's API, which kush never makes. You type the name you want and it re-pins.

Why kubectl and your prompt just follow along

kush ns only edits the namespace: field in the temp kubeconfig that KUBECONFIG already points at. kubectl re-reads that file on every invocation, so the switch lands with no flag, and starship's kubernetes module reads the same file, so your prompt updates on its own. Nothing anywhere caches the old namespace.

Edit this page