A shout-out to Radar

Radar, from Skyhook

Radar calls itself "the missing open source Kubernetes UI", and I think it's earned the name. It's a single binary you can run on your laptop against whatever cluster you're pointed at, or deploy in-cluster, and it needs no account and no cloud backend. You get a live topology graph, an event timeline, Helm releases, Argo CD and Flux, a best-practice audit and a built-in MCP server for AI tools. It keeps your cluster data local, and it's Apache 2.0.

I'd just started digging into Hangar when I saw Radar announced, on LinkedIn I think. I was living in Kubernetes at the time, and I love things that are pretty and functional in equal measure, so I gave it a try. I liked it straight away. It's the kind of interface I keep arguing platforms should have: calm, fast, and honest about what the cluster is doing.

The people behind it

Radar is built by Skyhook, a Y Combinator company (W23) founded by Oded Regev, its CEO, and Nadav Erell, its CTO. Nadav is also, by a wide margin, Radar's most prolific committer. What stood out to me as a contributor was how open they are: the maintainers reviewed my work carefully, merged it, and polished a few edges themselves on the way in.

Radar is their gift to the community. Their product is Skyhook itself, an internal developer platform out of the box: a PaaS-like experience on Kubernetes, running in your own cloud, so a team can ship production-ready services without first hiring a platform team to build one. That's the same problem Hangar is my answer to, and it's good to see it offered as a product people can simply pick up.

What I added, and why

My day-to-day was full of Tekton pipelines and Argo Rollouts, and Radar didn't have much for either yet. At the same time I was sharpening my skills with AI coding tools, so I set myself a question: how hard would it be to add the features I wanted to someone else's codebase, one I'd never seen before, to a standard its maintainers would merge?

Not as hard as I'd feared, and more fun than I'd expected. The AI did a lot of the typing. Knowing what a PipelineRun or an AnalysisRun should look like when things go wrong, testing against real clusters, and deciding what was worth a maintainer's time was still on me. Seven of my changes are merged so far, and the Tekton work is in review.

All of these live in skyhook-io/radar. Radar also supported Argo Rollouts before I arrived; my part was building on what was there.