The cluster already knows why the pod died. Polylane reads the events, lines them up with the rollout, and tells you.
Connect any cluster with one read-only ServiceAccount, and Polylane works out why pods crash, get evicted or never become ready.
Get started for freeEverything Polylane picks up.
The cluster objects Polylane reads.
Namespace Node Deployment StatefulSet DaemonSet ReplicaSet Job CronJob Pod Service Ingress ConfigMap Secret Service Account Role Role Binding Cluster Role Cluster Role Binding Persistent Volume Persistent Volume Claim Storage Class Network Policy Horizontal Pod Autoscaler Pod Disruption Budget The cluster knows. Now someone's listening.
Every workload and what it depends on
Deployments, StatefulSets, DaemonSets, pods, services, ingresses, CronJobs, volumes, autoscalers, roles and bindings all sit in one graph with their history, next to the cloud the cluster runs on. A failing pod leads straight to the node and the cloud underneath it.
Straight to the cluster
Polylane inspects pods and reads logs and events around the clock, judging what it sees against how each workload normally behaves. No dashboards in the loop.
The failures Kubernetes hides well
Crashloops, OOM kills, pods stuck pending and rollouts that never converge don't always page anyone. Polylane investigates each one and writes up what happened.
What changed before it broke
Polylane records what appeared, changed or disappeared in the cluster. When a workload starts failing right after a rollout, the investigation starts from that rollout.
get, list, watch, and nothing else
Connect with a kubeconfig or a read-only ServiceAccount token. Polylane only makes read-only Kubernetes API calls, credentials are encrypted before they're stored, the agent never sees the token, and you can scope it to a single namespace.
Fixes for code and manifests
When the root cause is a line of code or a manifest, Polylane writes the fix on the repo behind the workload, with the evidence attached.
These prompts already work. Copy one, swap your own service name in, and run it.
Give it to your coding agent and let it cook.
“If checkout-api goes down, what goes down with it? Walk the graph and rank by traffic.” 
“Find things that look wrong but never alerted: error spikes, dead crons, growing queues.” “Map our AWS account: what's running, what depends on what, and what's critical?” 
“Find the risky configuration: public buckets, missing alarms, single points of failure.” Use Kubernetes with any coding agent. Polylane connects them.
Connect your agent to Polylane once and it can query Kubernetes alongside the rest of your stack.
Guides and fixes for Kubernetes
- Error fix Kubernetes CrashLoopBackOff: causes and fixes What CrashLoopBackOff means in Kubernetes, how to read the exit code and events behind it, and the fix for each common cause, from bad config to failing probes.
- Error fix Kubernetes OOMKilled (exit code 137): causes and fixes Why Kubernetes reports OOMKilled with exit code 137, how to tell a limit kill from a node eviction or a plain SIGKILL, and how to size memory so it stops.
- Guide Is It Safe to Give an AI SRE Read Access to Production? Read access for an AI SRE is a sound first step if you scope it. The risks are secrets, customer data and exfiltration. Here are the controls and a checklist.
Common questions.
Is Polylane an AI SRE for Kubernetes?
It does that work, though we call it self-operating software. Polylane keeps your pods, deployments and nodes in one graph, investigates what breaks and fixes it with your approval, so the cluster looks after itself.
How does Polylane connect to my Kubernetes cluster?
Paste a kubeconfig, or enter the API server URL and a token. A read-only ServiceAccount with get, list and watch is all it needs, encrypted before it's stored.
Does it work with EKS, GKE, or my own clusters?
Any cluster that speaks the Kubernetes API: EKS, GKE, AKS, or bare metal. Connect an AWS account and its EKS clusters land in the same graph automatically.
What Kubernetes resources does Polylane sync?
Namespaces, nodes, Deployments, StatefulSets, DaemonSets, Jobs, CronJobs, pods, services, ingresses, ConfigMaps, secrets, RBAC, volumes, network policies, and autoscalers. The full list is on this page.
Can Polylane change my cluster?
No. Polylane only makes read-only Kubernetes API calls. Fixes go to the repos behind your workloads, for you to review and merge.
Does it work with my observability provider?
Yes: Datadog, Sentry, Honeycomb, Axiom, Grafana Cloud, Logfire, Better Stack and OpenStatus. Forward their alerts and Polylane triages each one.
What does it cost?
There's a free plan, and paid plans start at $80 a month. On Scale and above you can switch on pay-as-you-go beyond the included credits. Every limit is published at polylane.com/pricing.