Related reading: FreeLens vs OpenLens vs Lens: which one to standardize on.
If you’ve been following Kubernetes tooling, you’ve probably already been through the Lens saga: Lens went commercial, OpenLens emerged as the community fork, then FreeLens appeared when OpenLens maintenance slowed. The pattern is familiar — a useful desktop tool, a licensing decision, a fork, another fork.
Radar is not a fork. It’s a different approach to the same problem: giving engineers a useful interface for Kubernetes clusters without the friction of kubectl for every task. Built by Skyhook (YC-backed, Google Cloud Partner), it’s been live since 2025, has 1.7k+ GitHub stars, releases weekly, and the founder reaches out to the community directly. That’s usually a good signal that someone is genuinely building in public.
This article covers what Radar actually does, where it pulls ahead of OpenLens and FreeLens, and when those tools are still the right choice.
Before getting into Radar specifically, it’s worth naming the landscape clearly:
- Lens — the original. Electron-based, polished, now commercial (Mirantis). The free Personal tier is non-commercial only. Pro is ~$22-35/user/month.
- OpenLens — the community fork of Lens before Mirantis closed exec/logs/shell in v6.3 (January 2023). Maintenance has slowed significantly. No active release cadence.
- FreeLens — a more active community fork, filling the gap left by OpenLens’ decline. Restores the missing features. No commercial backing.
- k9s — terminal TUI, fast, keyboard-driven, single-cluster. Different audience.
- Headlamp — CNCF Sandbox project, plugin-extensible, web-based.
- Radar — Go binary, Apache 2.0, team-oriented, topology and event timeline focused.
The problem with OpenLens and FreeLens is not that they’re bad tools — they’re genuinely useful for the solo developer with one or two clusters. The problem is that they’re single-cluster-at-a-time desktop apps with no concept of team, no persistent state, and no awareness of the modern Kubernetes ecosystem (ArgoCD, Flux, Karpenter, KEDA). As your infrastructure grows, you outgrow them.
What Radar Actually Is
Radar is available in two forms:
- Radar OSS — a single ~30MB Go binary, Apache 2.0, free forever. Can run locally (desktop app) or deployed in-cluster via Helm. No sidecars, no feature gates.
- Radar Cloud — same binary, adds a hosted control plane with fleet aggregation, 30-day event retention, SSO/SCIM, scoped RBAC, and shared URLs for team incident response. Priced per cluster ($99/cluster/month for Team), not per user.
The per-cluster pricing is a deliberate design decision — teams don’t pay more as they add engineers, only as they add clusters. For a 20-person platform engineering team managing 5 clusters, Radar Cloud runs $495/month. The equivalent Lens Pro seats would cost $2,200-4,200/month.
For most self-hosted environments, the OSS version is sufficient and costs nothing.
Key Features
Topology View
This is the most visually distinctive feature. Radar renders a live service graph for your cluster: deployments, services, ingresses, cross-namespace dependencies, and east-west traffic flows — all in a single view without running kubectl get all -A and stitching the output together mentally.
OpenLens and FreeLens have resource list views. They show you what exists. Radar shows you how things connect — which is what you actually need when debugging why Service A can’t reach Service B.
Persistent Event Timeline
Kubernetes events are ephemeral by default — they expire after approximately one hour. When something breaks at 2am and you’re looking at it at 9am, the events that explain what happened are gone. Logs may still be there if you’re running a log aggregator, but the Kubernetes-level events (pod restarts, scheduling failures, node pressure events, probe failures) are gone.
Radar retains events. The OSS version extends this beyond the default 1-hour cluster retention. The Cloud version retains 30 days. You can rewind the timeline to any point and reconstruct what the cluster looked like at that moment.
Neither OpenLens nor FreeLens have any event retention beyond what the cluster itself provides.
GitOps Integration (ArgoCD + Flux)
Radar auto-detects ArgoCD and Flux and surfaces sync state, drift, and health directly in the UI. You can see whether a deployment is in sync, when it last synced, and whether it drifted from the desired state in Git.
In OpenLens and FreeLens, ArgoCD resources appear as generic Kubernetes custom resources. You can see the CRDs, but there’s no purpose-built understanding of what they mean — no sync status visualization, no diff view, no rollback trigger.
Helm Management
Radar tracks Helm releases with full revision history and supports one-click rollbacks from the UI. This is similar to what OpenLens/FreeLens offer via the Helm releases view, but Radar adds revision diffing — you can see what changed between release 5 and release 6 before deciding to roll back.
Image Filesystem
You can browse container image filesystems through Radar without needing kubectl exec into a running pod or access to the container registry. Useful for security audits and debugging — you can verify what’s actually in an image at rest.
MCP Server (AI Integration)
Radar ships with an MCP (Model Context Protocol) server, which means you can connect Claude, Cursor, or GitHub Copilot directly to your cluster context and ask questions about it in natural language. The MCP server is token-optimized — it doesn’t dump raw YAML at the model, it structures cluster state into meaningful context.
This is something neither OpenLens nor FreeLens have. It’s also something that’s genuinely useful if you’re already using AI assistants for development work.
Cluster Audit
30 built-in best-practice checks — resource requests/limits, RBAC permissions, image pinning, network policies, security contexts. The checks are labeled by compliance framework. This is not a replacement for dedicated security tooling (Trivy, Falco, Polaris), but it’s a useful first-pass audit without leaving the tool you’re already using.
Multi-Cluster Support (Cloud)
The Cloud tier adds fleet-level visibility: a single view across all clusters, cross-cluster search, and drift detection between environments (e.g., staging vs. production). This is the feature that changes the calculus for platform engineering teams managing 5+ clusters.
OpenLens and FreeLens require you to switch cluster context manually. There is no fleet view.
Architecture: Why a Go Binary Matters
OpenLens and FreeLens are Electron apps — Chromium + Node.js wrapped in a desktop shell. This means:
- 200-500MB install size
- 1-2 second startup time on a fast machine, more on slower ones
- Memory footprint in the hundreds of megabytes
- Local kubeconfig required on each engineer’s machine
Radar’s in-cluster deployment is a single Go binary (~30MB) that runs as a Pod with a ServiceAccount. It connects to the hosted control plane over outbound WebSocket + TLS. No inbound firewall rules, no kubeconfig distribution, no per-engineer setup.
The local desktop app is also a lightweight Go binary — 65-second startup was demonstrated on a 322-node cluster. That’s not a typo.
For in-cluster deployment, the architecture means security is handled at the ServiceAccount level, not by distributing kubeconfigs to engineer laptops. That matters for teams with security requirements around credential management.
Feature Comparison
| Feature |
Radar OSS |
Radar Cloud |
OpenLens |
FreeLens |
| License |
Apache 2.0 |
Proprietary (hosted) |
MIT/GPL |
MIT |
| Maintenance |
Active (weekly releases) |
Active |
Stalled |
Active (community) |
| Architecture |
Go binary / in-cluster |
In-cluster + hosted |
Electron |
Electron |
| Multi-cluster |
Basic |
Fleet view |
 |
|
| Event retention |
Extended |
30 days |
Cluster default (~1h) |
Cluster default (~1h) |
| Topology view |
 |
|
|
|
| GitOps (ArgoCD/Flux) |
|
|
CRDs only |
CRDs only |
| Helm management |
|
|
|
|
| kubectl exec / logs / shell |
|
|
(restored) |
|
| MCP / AI integration |
|
|
|
|
| Cluster audit |
|
|
|
|
| SSO / SCIM |
|
|
|
|
| Shared incident URLs |
|
|
|
|
| Image filesystem browser |
|
|
|
|
| Cost tracking |
(OpenCost) |
|
|
|
| Price |
Free |
$99/cluster/month |
Free |
Free |
When Radar Makes Sense
You’re managing multiple clusters. Even with the OSS version, the topology view and event timeline make Radar more useful than OpenLens/FreeLens at 3+ clusters. The Cloud fleet view is the compelling option at 5+.
Your team uses GitOps. If ArgoCD or Flux is part of your workflow, Radar’s native understanding of sync state and drift is meaningfully better than seeing CRDs in a generic list view.
You need post-mortem capability. If your incident review process involves looking at what the cluster was doing when the alert fired, you need event retention. Radar has it; OpenLens and FreeLens don’t.
You’re adopting AI tooling. The MCP server is the most forward-looking feature here. If you use Claude Code, Cursor, or Copilot for your infrastructure work, having cluster context available to those tools without copy-pasting YAML is a genuine productivity improvement.
You have a platform engineering team. Per-cluster pricing, SSO, SCIM, and shared incident URLs are features that only matter if you have more than one person managing infrastructure.
When OpenLens or FreeLens Still Makes Sense
You’re a solo developer with one or two clusters. OpenLens and FreeLens are familiar, local, and have zero setup overhead. If you don’t need team features, event retention, or topology views, they remain perfectly functional tools.
You’re deeply invested in the Lens UX. The resource tree, the terminal integration, the way Lens presents namespace-scoped resources — if your muscle memory is built around that interface, switching has a real cost. Radar is different, not just better.
You need maximum customi
