frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Kelos – Run Claude —dangerously-skip-permissions on Kubernetes

https://github.com/kelos-dev/kelos
2•gjkim042•1h ago
Kelos is a Kubernetes framework for orchestrating autonomous coding agents. It runs Claude Code, Codex, Gemini, OpenCode, or custom agents safely in isolated, ephemeral pods.

The idea is that you define your everyday development workflows as YAML and let them run continuously on Kubernetes. Agent tasks are Kubernetes CRDs, so things like:

- Watch for “bug” issues → auto-draft a fix PR

- Auto-review incoming pull requests

- Auto-triage new issues with labels and priority

- Periodically scan the codebase → propose improvements

- Test the project as a new user → surface rough edges

Kelos is using this on itself. The self-development pipeline picks up open issues, investigates them, opens or updates a PR, self-reviews, and retries if CI fails. When the agent gets stuck, it labels the issue needs-input and stops:

https://github.com/kelos-dev/kelos/tree/main/self-developmen...

I still do the final review and manually merge PRs, but it’s been quite helpful.

https://github.com/kelos-dev/kelos

Happy to answer questions about the design or what’s broken.

Comments

jlongo78•23m ago
running claude with --dangerously-skip-permissions in k8s is actually a smart move for isolation. the container boundary does alot of the permission work for you. one thing i'd watch though: ephemeral pods mean lost session context. the real workflow pain isnt the permissions, its picking up where you left off after a pod restart. stateless infra and stateful agent conversations dont mix well without something managing session persistence externally.