Kubernetes Management

Make Sense of Your Kubernetes Clusters

Connect the Kubernetes clusters you already run and instantly understand what’s deployed, what’s exposed, what’s failing, and what should be cleaned up, without changing your deployment workflow.

Inventory snapshot 2026-07-06 5:00 AM ▾
Diff with Previous
Namespace
payments-prod
Active
+ Added
Ingress
payments-edge
payments.example.com
+ Added
Service: ClusterIP
payments-api
10.96.20.30 · https TCP/443
+ Added
Deployment2 / 2 Ready
payments-api
apps/v1 · 2 replicas
+ Added
PodReady
payments-api-6f4c-2k8mn
gke-payments-pool-1
+ Added
PodReady
payments-api-6f4c-9qz4d
gke-payments-pool-2
+ Added
ConfigMap
payments-api-config
LOG_LEVEL · FEATURE_PAYMENTS_V2
+ Added
ConfigMap
legacy-flags
unused since May
Snapshots collected automatically — scrub through time, diff on demand
Runs anywhere your clusters do
AWSGoogle CloudAzure+ private cloud
One complete loop

Register, inspect, optimize, clean up

Hyphen Agent watches cluster state, surfaces changes, recommends fixes, and applies only the actions your team approves.

01

Register what you have

One kubectl apply installs Horizon, Hyphen's in-cluster agent. It connects the cluster to your organization and starts reporting inventory for private cloud, AWS, Google Cloud, and Azure clusters.

$ kubectl apply -f hyphen-horizon.yaml
namespace/hyphen-horizon created
✓ prod-east connected; inventory reporting
02

See what changed

Inventory snapshots capture cluster state over time. Diffs show added, removed, and changed resources, including replica changes, so a weekly review does not turn into a kubectl search.

Snapshot #142 · diff with previous
12m ago
+ 6 added− 3 removed~ 2 changed
payments/api replicas 12 → 8 · ready 8/8
03

Optimize with approval

Ask Agent to review a workload before a launch, after an incident, or during normal review. It recommends replica changes backed by evidence, skips HPA-managed workloads, and rechecks state before applying what you approve.

Scale checkout/web 6 → 10 before launch-nov-12
Approve
Planned event raises expected demand. Readiness stable at 6/6; no HPA detected.
04

Clean up on a schedule

A cleanup policy scans for stale resources on a schedule. Use per-cluster overrides for cadence, thresholds, and ignored namespaces. Scans stay read-only until you approve a deletion.

14
stale candidates found this scan; awaiting decisions
weekly
policy override on legacy-dc · 30-day threshold
prod-east
Ready
AWSAWS · v1.31.2
● First inventory collected 2 minutes after install
12Namespaces
48Workloads
214Pods
36Services
9Ingresses
Discovered automatically. Nothing was redeployed.
Existing clusters

Works with the clusters you already run

You do not need to deploy through Hyphen first. Point the agent at an existing production cluster and it inventories the resources already running, including namespaces, workloads, services, and config. No migration, redeploy, or downtime.

Visibility

See what’s running and what needs attention

Each namespace is shown as a snapshot of ingresses, services, workloads, pods, and config. Not-ready workloads surface quickly, and warnings are grouped by reason so the team can start with the affected component.

Nodes3 / 3 ready
Workloads9 workloads in 2 apps1 NOT READY
Pods6 ready1 NOT READY
Exposure1 public LB
Warnings28 warnings
Namespace
payments-prod
Active
Ingress
payments-edge
payments.example.com
Service: ClusterIP
payments-api
app: payments-api
10.96.20.30
https TCP/443
Deployment
payments-api
1 / 2 Ready
PodRunning · Ready
payments-api-6f4c7f9b74-2k8mn
gke-payments-pool-1
PodCrashLoopBackOff · Not Ready
payments-api-6f4c7f9b74-9qz4d
restarts: 14 · gke-payments-pool-2
ConfigMap
payments-api-config
LOG_LEVEL · FEATURE_PAYMENTS_V2
Kubernetes
+ Cluster
Name
Status
Cloud Provider
prod-east
Ready
AWSAWS
gcp-payments
Ready
Google CloudGoogle Cloud
azure-analytics
Ready
AzureAzure
on-prem-lab
Cleanup needed
Private
Fleet

Every cluster in one console

EKS, GKE, AKS, and private clusters appear in one list. Provider, state, and last-seen status are detected automatically, so the team has a shared view of which clusters exist and whether they are healthy.

Agent

Optimization and cleanup with approval built in

Hyphen Agent scans each cluster on your schedule, recommends workload changes, and flags resources idle past your threshold, such as abandoned deployments, orphaned config, and completed jobs nobody deleted. Nothing changes in your cluster until someone on your team approves it.

Cleanup Policy ⓘ
Enable Kubernetes cleanup
Turn this off to disable scheduled Kubernetes cleanup only for this cluster
Schedule
Every day ▾
Cleanup Threshold
Days before an idle resource is flagged for cleanup
14
Ignored namespaces
kube-system
kube-public
hyphen-horizon
Pending recommendations
2 awaiting approval
deployment/legacy-report-genIdle 21 days · scale to zero
ApproveDismiss
configmap/old-feature-flagsUnreferenced · remove
ApproveDismiss
No changes are made to your cluster without approval.

Frequently Asked Questions

One kubectl apply installs Horizon, Hyphen's in-cluster agent. It connects the cluster to your organization and starts reporting inventory—no migration, no redeploy, and no changes to how your team ships. You do not need to deploy through Hyphen first; point it at the clusters you already run on AWS, Google Cloud, Azure, or private cloud.
No changes are made to your cluster without approval. Cleanup scans stay read-only until you approve a deletion, and optimization recommendations are applied only after your team signs off. Before applying an approved change, the Agent rechecks cluster state to make sure the recommendation is still valid.
Inventory snapshots capture cluster state over time, and diffs show added, removed, and changed resources—including replica changes—so a weekly review or post-incident investigation doesn't turn into a kubectl archaeology session. You can see what's deployed, what's exposed, and what's failing across namespaces, workloads, pods, services, and ingresses.
A cleanup policy scans for stale resources on a schedule you control, with per-cluster overrides for cadence, idle-day thresholds, and ignored namespaces. Scans are read-only: resources are flagged for review, and nothing is deleted until you approve it. You can disable scheduled cleanup per cluster at any time.
Yes. When you ask the Agent to review a workload—before a launch, after an incident, or during routine review—it recommends replica changes backed by evidence and skips HPA-managed workloads entirely, so it never fights your autoscaler.
Yes. Hyphen puts every cluster in one console regardless of where it runs—AWS, Google Cloud, Azure, or private cloud—with a fleet view of status, provider, and inventory so you're not context-switching between cloud consoles to answer basic questions.
Minutes. Once Horizon is installed with a single kubectl apply, the cluster starts reporting inventory and appears in your console. You can start free with no credit card required.

Bring your clusters into focus

Connect your first cluster in minutes. Start free, no card required.