Hey everyone,
I've been working on a project called KubeAgent for the last few months. Honestly, it started because I was tired of waking up to a down service or a CrashLoops on my personal clusters at ungodly hours. it always happens when you don't look. the pressure was intense and most AI tools I tried were useless.
why?
I didn't want to build another heavy SaaS with a million dashboards. I just wanted something that:
- Dead simple and easy to setup
- Runs locally in my terminal (no cluster access for us, stay paranoid!)
- Actually tries to fix the problem (restart, scale, etc.)
- Asks me via Slack before doing anything "risky" (this bit is still in testing. might not work reliably)
It's still early and definitely has some rough edges (working on the integrations right now), but it's finally at a point where it's saving me hours and gives me a piece of mind.
how it works?
it has a KB system that learns about your cluster and your codebase(s). because it knows how everything works in your setup, the fixes are more reliable. all safe actions are done automatically and you are in charge of risky stuff. it pings you on Slack and Telegram if something goes wrong and will ask for permission if an approval is needed.
I would love to hear what you guys think about the "Zero-Access" and the "Human-in-the-Loop" approach. Is it too minimal? Just right?
Happy to answer any technical questions about the agentic loop too!
Top comments (1)
the website: kubeagent.net/