Local binary connects directly to the k8s cluster (queries the API server) using the same Go SDK as k8s itself. Keep a local cache and listens to live updates using k8s native watchers.
The load on the server is not that much for a reasonable number of users, each just syncs the data to their local client.
It also supports install with Helm in the cluster, so a single installation can serve any number of users - data gets synced once (continuously), every user gets served the local embedded UI. You probably want to integrate with auth for that - supports proxy and OIDC natively.
Creator here. We built Radar because the Kubernetes dashboard/IDE space got frustrating:
- Lens got acquired, went partially closed-source, became slow, now requires login
- k9s is solid but terminal-only with a steep learning curve
- Official Kubernetes Dashboard is archived (Headlamp is the successor but missing features and not amazing UX)
- Enterprise tools want per-node pricing and mandatory cloud sync
We had built internal tooling at Skyhook for debugging clusters. Customers kept asking for it, so we carved it out and open-sourced it.
Radar is local-first: single binary, runs on your machine (or can install on cluster if you prefer), uses your Kubeconfig, blazing fast. No agents, no CRDs, no account. Works airgapped.
It also supports install with Helm in the cluster, so a single installation can serve any number of users - data gets synced once (continuously), every user gets served the local embedded UI. You probably want to integrate with auth for that - supports proxy and OIDC natively.