Radar - 2026 Pricing, Features, Reviews & Alternatives

Kubernetes is easy to query. Hard to understand. Radar is the missing Kubernetes interface for humans and agents — an open-source tool that makes a live cluster make sense over two surfaces: a UI optimized for humans, and an MCP server optimized for AI agents.

Radar runs as a single Go binary that talks to the Kubernetes API directly. Run it locally on your laptop with zero cluster-side install, or deploy it in-cluster via Helm for shared team access. There are no CRDs to apply, and in local mode no account is required. It works against any conformant cluster — GKE, EKS, AKS, minikube, kind, or k3s — is airgapped-ready, and makes no external calls. Radar watches your cluster through informers and streams live updates to the browser, so what you see is always current.

Two surfaces, one engine:

FOR HUMANS — a fast, real-time UI:

• Topology — a live resource graph, auto-laid-out, with a network-traffic mode.

• Resource explorer — full CRD auto-discovery, status/problem filtering (CrashLoopBackOff, ImagePullBackOff), one-click YAML, logs, events, and related resources.

• Event timeline — every Kubernetes event and delta in one real-time feed, retained past the 1-hour TTL.

• Log streaming & pod exec — tail logs and open a shell from the UI.

• Image filesystem viewer — browse any container image's filesystem, no kubectl exec or docker needed.

• Helm — inspect values, compare revisions, upgrade, roll back, or uninstall.

• GitOps — first-class FluxCD and ArgoCD support with sync status, health, and reconciliation controls.

• Cluster audit — 31 best-practice checks across security, reliability, and efficiency, framework-labeled (NSA/CISA, CIS), each with remediation guidance.

• TLS certificate & cost insights — catch expiring certs before outages; see cluster and workload spend via OpenCost.

FOR AI AGENTS — a built-in MCP server:

• Claude, Cursor, Copilot, and other assistants query your cluster through Radar over the Model Context Protocol.

• Data is pre-processed and token-optimized instead of raw YAML, and read tools are strictly read-only by default — AI gets useful cluster context, safely.

Radar is and always will be open source. The engine is Apache 2.0, self-hostable with no node or cluster limits, and free forever — no feature gates, no license keys, no "call us for pricing" features. Run it locally per-developer, or deploy in-cluster via Helm for shared access with OIDC/proxy auth and per-user Kubernetes RBAC.

Radar Cloud is the hosted, team-scale layer on the same engine — for the things a single binary can't reasonably do, not because the OSS is crippled. The same binary runs as a lightweight agent in your clusters and dials out to the control plane, bringing every cluster behind one URL. Cloud adds fleet aggregation (audit failures, package drift, cross-cluster resource search, expiring certs), centralized SSO (Google, GitHub, SAML/OIDC), organizations and scoped RBAC roles, persistent audit-log retention with SIEM export, saved investigations and shareable incident snapshots, and routed alerts via Slack, PagerDuty, MS Teams, and webhooks. Cloud is priced per cluster, never per seat — a free tier for 3 clusters, Team at $99/cluster/month, and Enterprise with SCIM, BYOC/on-prem, and a 99.9% SLA. SOC 2 Type 2 applies to every plan.

Install in about 30 seconds.

Open source on GitHub (search for Radar K8s)