Radar
Open-source Kubernetes UI with an MCP server so on-call engineers and AI agents read the same live cluster.
Radar is the rare Kubernetes tool where the free build isn't a demo — MCP, topology, event timeline, Helm, image filesystem, 44 audit checks, and per-cluster cost all ship under Apache 2.0. The agent angle is real: v1.15.0 put cost and rightsizing in front of MCP clients, and Skyhook's own benchmark has an agent diagnosing faults 3x faster on Radar than on plain kubectl. Pay for Cloud only when you want alerting around the clock or one link to every cluster.
Verified 2d ago · liveness 81/100 · cite: rightaichoice.com/tools/radar
- Platform engineers running multi-cluster fleets who want one search bar across every cluster
- On-call teams that want an AI agent investigating read-only while they read the topology graph
- DevOps teams on Helm plus Argo CD or Flux who want drift, revision state, and manifest diffs in one screen
- Platform teams that need SSO, scoped RBAC, and an audit log of privileged cluster actions
- Engineers who want a terminal-only workflow and never open a web UI — stick with k9s
- Teams needing APM or in-service distributed tracing; Radar watches Kubernetes, not your code
- Shops looking to replace Grafana or Datadog wholesale — network visibility here is K8s-scoped
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Radar if you need application-level APM or in-service distributed tracing, or if you want a terminal-only workflow with no graph and no timeline.
Team and Enterprise include 100 and 200 investigations per cluster per month respectively — past that each additional investigation is $1, opt-in.
OSS is free forever with unlimited clusters and nodes, so single-cluster teams pay $0. Radar Cloud Team is $149 per cluster/month and Enterprise $299 per cluster/month, billed per cluster rather than per seat, with 20% off on annual commitments and a 14-day trial on paid plans. Against per-seat tools like Lens that punish you for letting teammates look, Radar stays cheap at large headcounts and gets expensive past roughly 15-20 clusters, where a volume Custom contract starts to make sense.
In short
Radar — Open-source Kubernetes UI with an MCP server so on-call engineers and AI agents read the same live cluster. Best for Platform engineers running multi-cluster fleets who want one search bar across every cluster, On-call teams that want an AI agent investigating read-only while they read the topology graph, DevOps teams on Helm plus Argo CD or Flux who want drift, revision state, and manifest diffs in one screen. Free to start; paid plans from $149/mo.
What's new in Radar
Checked todayAcross the latest 5 updates: 4 feature updates and 1 changelog entry.
OSS v1.16.0 — Issues: one steady issue per crash loop
A crash-looping container is now one issue until it recovers, with probe failures, last exit, and looping replica counts attached as evidence. Severity follows impact and the issue clears about ten minutes after the crash stops.
OSS v1.16.0 — Log viewer: the right levels, whole stack traces, and Hide matching
Levels now come from what the logging library wrote — JSON and logfmt fields, controller log headers, common frameworks — rather than keyword guessing. Filtering to ERR keeps the whole stack trace, and Hide matching drops lines like grep -v.
Radar Cloud: your Kubernetes RBAC decides what people and AI can do
Hub roles cover Radar's own features while Kubernetes RBAC decides per-person cluster access, including IdP-group bindings. An AI investigation reads the cluster as the person who started it, and background alert investigations use a view-only radar:ai:reader group. Helm secret v
Radar v1.15.0 exposes rightsizing and cost data to MCP agents
Two read-only MCP tools add cost (namespace, workload, node breakdowns and trends) and CPU/memory rightsizing data. Risk signals such as OOM, limit conflicts and throttling surface before cost savings.
Radar v1.15.0 GitOps: full manifest diffs and correct destination clusters
Drifted Argo CD resources can be diffed side-by-side or unified against live manifests, and remote resources now report which cluster they live on with Open in <cluster>.
What people actually say about Radar — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
66 mentions across 4 sources (Hacker News, App Store, GitHub, Lemmy) · researched Jul 2, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Live topology graph with SSE updates for real-time cluster visibility.
- +Container image filesystem browser eliminates kubectl exec for debugging.
- +Built-in Helm release tracking, comparison, and rollback capabilities.
- +GitOps visibility for ArgoCD and Flux in one dashboard.
- +MCP server enables AI agents like Claude to interact with the cluster.
- −Almost no real user reviews exist for the Kubernetes tool.
- −Free tier event retention is only 1 hour, limiting debugging scope.
- −Self-hosted version lacks multi-cluster fleet management features.
- −Cloud layer needed for SSO, persistent retention, and alerts.
- −No community benchmarks or uptime data for reliability validation.
- • Cloud tier pricing undisclosed — may scale with cluster count
- • Self-hosted requires infrastructure for running the tool in-cluster
Viability Score
How well maintained and how widely used is Radar? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Live resource topology graph, ELK.js-laid-out, streamed over SSE
- Persistent event timeline retaining K8s events past the default 1-hour TTL
- Issues: one steady issue per crash loop with probe failures, last exit, and looping replica counts
- Failures folded to root cause with blast radius
- Log viewer with library-derived levels, whole stack traces, and Hide matching
- MCP server for AI agents — read-only, keyless, token-optimized
- Bring your own agent CLI: Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Google Antigravity
- Cost and rightsizing exposed to MCP agents (namespace, workload, node breakdowns and trends)
- Risk signals before savings: OOM, limit conflicts, throttling
- Hosted AI investigations that run without a local CLI (Cloud plans)
- GitOps: Argo CD and Flux drift with full manifest diffs against live resources
- Remote GitOps resources report their destination cluster with Open in <cluster>
- Argo Rollouts operations: promote, promote fully, skip, abort, or retry, permission-gated
- Network reachability tracing declared path and live-probing DNS, TCP, TLS, and HTTP
- Cluster audit — 44 best-practice checks, framework-labeled
About Radar
Radar is an Apache 2.0 Kubernetes visibility tool from Skyhook that ships as a single binary you install with Homebrew, Helm, Krew, Scoop, or a curl script, and run on your laptop or in your cluster without creating an account. The same engine drives two surfaces: a web UI written for humans, and an MCP server written for AI agents like Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, and Google Antigravity. The repo sits at 3.8k stars with 73,000+ GitHub downloads in 2026. The human side folds raw workloads into application views with health, environments, and version skew, then backs them with a live resource topology graph laid out with ELK.js and streamed over SSE. A persistent event timeline keeps every Kubernetes event and delta past the cluster's default one-hour TTL, so you can rewind a resource to the hour before an incident. Around that core sit Helm release tracking with revision comparison and one-click rollback, a container image filesystem browser, Argo CD and Flux drift diagnosis with side-by-side or unified manifest diffs against live resources, a 44-check framework-labeled cluster audit, and network reachability tracing that declares the path, then probes DNS, TCP, TLS, and HTTP. The agent side is what separates Radar from the dashboard crowd. Its MCP server lets an agent investigate read-only and keyless, and v1.15.0 extended that surface with two read-only MCP tools covering namespace, workload, and node cost breakdowns, trends, and CPU/memory rightsizing data, with risk signals like OOM, limit conflicts, and throttling surfaced before the savings. Radar Cloud is the hosted layer, priced per cluster and never per seat: Free for up to three clusters, Team at $149 per cluster/month, Enterprise at $299 per cluster/month, plus Custom for air-gapped and regulated deployments. Cloud adds routed alerts, hosted investigations that run without a local CLI, fleet aggregation, longer retention, SSO and SCIM, and an SLA. If your team is weighing Lens
Behind the Verdict
Here's the split that matters. If you run one cluster, or five, and you want a UI that makes Kubernetes legible without a sales call, install Radar OSS and stop reading. Topology, event timeline past the one-hour TTL, Helm revisions with rollback, image filesystem, 44 framework-labeled checks — that's the open-source build, not a trial. The reason to consider Cloud isn't feature gating, it's the 2am problem. Free covers three clusters with 100 hosted investigations a month, 24-hour timeline retention, and alerts routed to Slack or any webhook. Team at $149 per cluster/month raises that to 100 investigations per cluster, 3-day timeline, 30-day audit, CEL-based alert rules, and a fleet-wide cost rollup. Enterprise at $299 adds SSO and SCIM, 200 investigations, 30-day timeline, and a year of audit retention with SIEM export. Per cluster, not per seat, is the sharpest thing about the pricing. Invite thirty engineers and the bill doesn't move. That's exactly the line Lens crossed for a lot of people, and it's why so many of the testimonials mention leaving Lens free. Where Radar bites: it watches Kubernetes and only Kubernetes. No APM, no in-service distributed tracing, no cloud spend outside your clusters. If you need Datadog or Grafana replaced wholesale, this isn't that and doesn't claim to be. And terminal purists who live in k9s will find Radar's value mostly on the graph and timeline side rather than in command speed. The v1.16.0 issue model is a quiet win for on-call. A crash-looping container becomes one steady issue with probe failures, last exit, and looping replica counts attached, instead of a fresh alert per restart — that alone cuts alert fatigue on noisy deployments. Combine it with the RBAC change in Cloud, where Kubernetes RBAC decides what a person and
Researching Radar? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Radar actually fits — and what changes day-one when you adopt it.
A checkout deployment is crash-looping in one namespace and nobody knows which cluster it lives on. You open Radar, search across every cluster from one bar, land on the crash loop, and see it as a single steady issue with probe failures and the last exit code attached instead of a fresh alert on every readiness flip.
Outcome: You rewind the event timeline to the hour before the first crash and find the config change that preceded it, then hand the shared link to the incident channel.
You point Claude Code or OpenCode at Radar's MCP server, which is read-only and keyless, and give it a live cluster. The agent pulls topology, events, and — now on v1.15.0 — namespace and workload cost and rightsizing data, with OOM and throttling risk signals flagged before it recommends any savings.
Outcome: The agent returns a diagnosis linked to Radar's evidence, on the same cluster view you're reading, and you keep control of anything that changes.
A resource is drifted. You expand the Argo CD resource in Radar and compare the manifest Git rendered against the one that's live, side by side or unified, with unchanged regions collapsed. A resource the Application deploys elsewhere now reports its destination cluster instead of opening a same-named object locally.
Outcome: You see exactly what Git wants and what's running, on the cluster it actually lives on, and either revert or commit the change.
Use Cases
- Debug a crashing pod by searching across every cluster, jumping to logs, and rewinding the event timeline.
- Hand Claude, Codex, Cursor, or OpenCode a read-only MCP connection so an agent investigates while you read the topology graph.
- Promote, abort, or retry an Argo Rollouts canary or blue-green progression from the Rollout page.
- Roll back a Helm release to a previous revision with one click and diff what changed between revisions.
- Diff a drifted Argo CD manifest against the live resource side by side or unified.
- Probe a service's HTTP endpoint from inside the cluster when port-forward is unavailable.
- Check whether a cluster is ready for the next Kubernetes upgrade, including removed gates and kubelet flags.
- Let an MCP agent pull namespace, workload, and node cost and rightsizing data before you resize requests.
Models Under the Hood
as of 2026-10-18
Limitations
- Radar OSS is Apache 2.0 and free with unlimited clusters and nodes, but the open-source build works one cluster at a time; Radar Cloud is needed for alerts, AI investigations, and a shared fleet view across clusters.
- Cloud starts free with up to three hosted clusters, and paid plans charge per cluster rather than per seat, with Self-Managed deployment limited to Enterprise and Custom.
- AI investigations rely on you bringing your own agent CLI (Claude Code, Codex, Cursor, etc.) on OSS, while hosted investigations are a Cloud capability.
as of 2026-10-09
Verification history
We have re-verified Radar 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Radar tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open-source Radar
$0
Ideal for
Homelab operators and single-team platform engineers who want the full engine without an account, a cloud dependency, or a credit card.
What this tier adds
Starting point: Apache 2.0, unlimited clusters and nodes one cluster at a time, full topology, Helm, image filesystem, 36-check audit, and MCP.
Cloud Free
$0
Ideal for
A small team that wants fleet visibility and Slack alerting across a handful of clusters before committing budget.
What this tier adds
Adds hosted fleet view, cross-cluster search, shared links, alerts, and 100 hosted investigations per month, capped at three clusters with 24-hour timeline retention.
Team
$149 per cluster/mo
Ideal for
Engineering teams running real production workloads across more than three clusters who need scoped access and longer history.
What this tier adds
Removes the three-cluster cap, lifts investigations to 100 per cluster/month, and adds scoped RBAC, 3-day timeline, 30-day audit retention, CEL alert rules, and fleet-wide cost rollup.
Enterprise
$299 per cluster/mo
Ideal for
Regulated or governance-heavy organizations that need SSO, SCIM, SIEM export, and the option to run Radar in their own infrastructure.
What this tier adds
Adds SAML/OIDC SSO and SCIM 2.0, 200 investigations per cluster/month, 30-day timeline retention, 1-year audit retention with SIEM export, Self-Managed deployment, and priority support.
Custom
Custom
Ideal for
Air-gapped, regulated, or very large fleets that need volume pricing, custom identity, data residency, or a contractual uptime guarantee.
What this tier adds
Adds volume pricing, custom IdP and data residency, custom AI usage and model configuration, custom retention, air-gapped deployment, a 99.9% SLA, and a dedicated CSM.
Where the pricing makes sense
The company stage and team size where Radar's pricing actually pencils out — and where peers do it cheaper.
OSS is free forever with unlimited clusters and nodes, so single-cluster teams pay $0. Radar Cloud Team is $149 per cluster/month and Enterprise $299 per cluster/month, billed per cluster rather than per seat, with 20% off on annual commitments and a 14-day trial on paid plans. Against per-seat tools like Lens that punish you for letting teammates look, Radar stays cheap at large headcounts and gets expensive past roughly 15-20 clusters, where a volume Custom contract starts to make sense.
Setup time & first value
How long it actually takes to get something useful out of Radar — broken out by persona, not the marketing-page minute.
OSS: one command — curl the install script or run the Helm chart — and the docs quote about 30 seconds to a live view, with no account or credit card. Cloud: sign up, install the Helm chart in each cluster, and connect it, which takes a few minutes per cluster including a Google or GitHub login. Enterprise Self-Managed adds the time to stand the control plane up in your own infrastructure. Agent
Switching to or from Radar
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Lens Desktop: point Radar at the same kubeconfig — the per-cluster cloud model replaces per-seat licensing, and the topology and event timeline cover what Lens free gated.
- →From k9s: keep the terminal for single-pod poking, add Radar when you need a graph, a retained timeline, or fleet-wide search across clusters.
- →From a Grafana-only stack: connect Prometheus so Radar can read Grafana Beyla and Alloy traffic, then use the traffic view for K8s-scoped path and policy questions.
- →From raw kubectl plus Argo CD UI: install Radar in-cluster beside Argo CD to get drift diagnosis, manifest diffs, and Rollouts operations in the same place.
- →From a Grafana or Beyla-only network view: add Calico policy intent reading so you can separate what is declared as allowed from what actually flows.
- ↗To k9s: if you never want a browser, drop the in-cluster Helm install and go back to a terminal UI against the same kubeconfig.
- ↗To Datadog or Grafana with tracing: Radar gives K8s-scoped network visibility only, so in-service APM and distributed tracing need a separate tool.
- ↗To Lens Desktop: if you want a desktop app with built-in cluster management and don't need MCP or a retained event timeline.
- ↗To a cloud-cost platform: Radar's cost rollup is cluster-scoped, so non-Kubernetes cloud spend still requires its own tool.
Integrations
Resources & Guides
- Documentationradarhq.io
Docs · Radar
Full product docs from radarhq.io
- Documentationradarhq.io
Llms · Radar
Full product docs from radarhq.io
- Resourceradarhq.io
Changelog · Radar
Helpful link from radarhq.io
- Resourceradarhq.io
Blog · Radar
Helpful link from radarhq.io
- Resourcegithub.com
Radar · Radar
Helpful link from github.com
Tutorials & Learning
YouTube returned 6 videos for “Radar”, and we withheld 6: 6 could not be judged, because “Radar” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Radar.
Official links
Featured Head-to-Head Comparisons
Radar vs Spider Cloud
Spider Cloud and Radar serve completely different needs. Spider Cloud is the right choice if you need fast, cost-effective web data extraction for AI agents and RAG pipelines—its new Browser AI commands and 1,000+ scraper catalog make it powerful. Radar is essential if you manage Kubernetes clusters and need a modern UI with live topology, Helm tracking, GitOps visibility, and an MCP server for AI-assisted debugging. Choose based on your problem: web data or K8s operations.
Radar vs Voyage Ai
Voyage AI is the clear choice for teams needing high-accuracy, domain-specific embeddings for enterprise RAG, especially in finance or legal. Radar serves a completely different need: open-source Kubernetes debugging and observability. Your decision depends entirely on whether you're optimizing search retrieval or managing clusters.
Radar vs Temporal Ai
Choose Temporal if you need a battle-tested durable execution platform for AI agents, microservices, or long-running workflows that survive failures. Choose Radar if you're a platform engineer or on-call team that needs a fast, open-source Kubernetes UI with live topology, Helm/GitOps insights, and AI-assisted debugging. They solve orthogonal problems — the right choice depends on whether your primary pain is workflow reliability or cluster visibility.
Popular in Developer Infrastructure
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
Frequently Asked Questions
Used Radar? Help shape our editorial sentiment research.