AI-powered diagnosis
Contextual fixes generated from events, logs, and manifest. Bring your own key (OpenAI, Anthropic, Groq, Gemini, GitHub Copilot) or run fully local with Ollama.
Kure Monitor
Stop debugging Kubernetes failures manually. Kure detects pod crashes, image-pull errors, and scheduling issues in seconds — and uses AI to tell you why.
Why Kure Monitor
Section titled “Why Kure Monitor”Most Kubernetes monitoring tools tell you that something broke. Kure tells you why, with context pulled from pod events, logs, and the manifest — analyzed by an LLM of your choice (OpenAI, Anthropic, Groq, Gemini, GitHub Copilot, or fully local with Ollama).
It also continuously audits your cluster for security misconfigurations and renders a live topology graph, all from a single dashboard.
Real-time failure detection
Catch CrashLoopBackOff, ImagePullBackOff, OOMKilled, FailedScheduling, FailedMount, and stuck Pending pods the moment they happen.
50+ security checks
Privileged containers, host namespaces, missing seccomp/AppArmor, root containers, untrusted registries, missing limits, RBAC misconfigs.
Topology diagram
Interactive graph of every workload, service, ingress, HPA, and NetworkPolicy. Click a node for the manifest, click an edge to focus that path.
Mirror pod testing
Deploy a temporary copy of a failing pod with the AI fix applied. Verify before committing to git.
Air-gapped friendly
Pair with Ollama and your cluster data never leaves your network. No telemetry, no SaaS dependency.
Install in two commands
Section titled “Install in two commands”helm repo add kure-monitor https://nan0c0de.github.io/kure-monitor/helm install kure-monitor kure-monitor/kure \ --namespace kure-system --create-namespace \ --set postgresql.password="$(openssl rand -hex 24)"Then port-forward and create the initial admin account:
kubectl port-forward svc/kure-monitor-frontend 8080:8080 -n kure-system# open http://localhost:8080Configure your LLM provider from the Admin panel after the first sign-in.
Full installation guide Helm, raw manifests, EKS / GKE / AKS, Ingress, LoadBalancer, and post-install checks.
Where Kure fits
Section titled “Where Kure fits”Kure is focused on failure diagnosis and security posture, not general observability. It complements Prometheus / Grafana / Datadog — it doesn’t replace them.
| Kure does | Kure does not |
|---|---|
| Real-time pod failure detection with AI fixes | Time-series metrics or alerting rules |
| 50+ security misconfiguration checks | Log aggregation or indexing |
| Live topology graph + mirror pod testing | Multi-cluster monitoring |
| Slack / Teams notifications on failure | Replace Prometheus or Datadog |
Open source, runs in your cluster
Section titled “Open source, runs in your cluster”Apache 2.0 licensed. Ships as a Helm chart. Stores nothing outside your cluster except whatever LLM provider you choose to call. The backend ServiceAccount is intentionally not granted read access to Secrets — secret values are never read.
See it in action Screenshots and short clips of the dashboard, AI troubleshooting, security scanning, and mirror-pod testing.
Read the docs Architecture, configuration, API reference, and troubleshooting.
Source on GitHub Star the repo, file an issue, or contribute.
Helm chart on Artifact Hub Browse versions and chart values.
Latest release notes What's new in 2.4.2 — sixteen new AI Advice detectors (38 total), Active/Ignored tabs for findings, and a fix for the false-positive Secret reference check.