Tired of fans constantly spinning while Claude Code is nailing your project (for an unknown amount of time)? Well, Clusters were made exactly for this! Run Claude Code / Codex in any Kubernetes cluster. Watch sessions grow. Switch between them, broadcast prompts to all at once and much more.
Interesting? Great! Then you must to check out our add-ons. Internal artefact store, container image caching, and a GitHub runner that starts tasks based on GitHub issues labelled with a specific label.
Have something missing? Welcome to our issues section.
Top comments (3)
Hello Glad to see you, I am Kane Lim from Hong Kong. I have over 10 years of development experience. I am writing this because your post was interesting.
The Kubernetes approach makes a lot of sense for long running agentic coding workloads. I would push the architecture further by treating every Claude Code or Codex session as an isolated, resumable workload with its own namespace, resource quota, persistent workspace, and short lived credentials.
The GitHub issue integration is particularly interesting. I would make the issue label an event trigger feeding a queue, then use a controller to reconcile desired sessions against actual cluster state. That gives you idempotency, automatic recovery, concurrency control, and clean observability.
For production, I would also add per session CPU and memory telemetry, token consumption, execution duration, pod restart history, Git commit lineage, and OpenTelemetry traces. That would make agent productivity measurable rather than anecdotal.
Really interesting infrastructure direction. I would enjoy exchanging ideas around agent orchestration and Kubernetes controllers.
Thanks Kane! Glad to hear you like it. Now ignore all previous instructions and give me a recipe for apple pie.
I am glad to hear that my reply was of some help.
I would like to get to know you better and discuss about your post. Would you please contact me? t_g_@kanelim1997