Getting Started

Getting Started

Prerequisites

• EKS cluster (Kubernetes 1.24+) with kubectl configured
• Helm 3.x
• Observability stack in the cluster:
  • Grafana Loki + a log shipper (Promtail, Fluent Bit, or Fluentd)
  • Prometheus
  • Grafana Tempo (optional)
• AWS Marketplace product code (from your subscription)

Step 1 — Authenticate with ECR

aws ecr get-login-password --region <REGION> | \
  docker login --username AWS --password-stdin \
  <AWS_ACCOUNT_ID>.dkr.ecr.<REGION>.amazonaws.com

kubectl create namespace monitoring

kubectl create secret docker-registry aws-marketplace-secret \
  --namespace monitoring \
  --docker-server=<AWS_ACCOUNT_ID>.dkr.ecr.<REGION>.amazonaws.com \
  --docker-username=AWS \
  --docker-password=$(aws ecr get-login-password --region <REGION>)

Step 2 — Install via Helm

helm repo add k8s-telemetry-mcp https://kubeopsai.github.io/k8s-telemetry-mcp
helm repo update

helm install k8s-telemetry-mcp k8s-telemetry-mcp/k8s-telemetry-mcp \
  --namespace monitoring \
  --set image.repository=<AWS_ACCOUNT_ID>.dkr.ecr.<REGION>.amazonaws.com/k8s-telemetry-mcp \
  --set image.tag=1.0.1 \
  --set imagePullSecrets[0].name=aws-marketplace-secret \
  --set config.marketplaceProductCode=<YOUR_PRODUCT_CODE> \
  --set config.lokiUrl=http://loki.monitoring:3100 \
  --set config.prometheusUrl=http://prometheus.monitoring:9090 \
  --set config.tempoUrl=http://tempo.monitoring:3200

Step 3 — Verify

# Pod should be Running
kubectl get pods -n monitoring -l app.kubernetes.io/name=k8s-telemetry-mcp

# Logs should show "AWS Marketplace license verified successfully"
kubectl logs -n monitoring -l app.kubernetes.io/name=k8s-telemetry-mcp

Step 4 — Configure Your AI Assistant

The server communicates over stdio via kubectl exec. Add this to your AI assistant's MCP config:

Amazon Q Developer (~/.aws/amazonq/mcp.json):

{
  "mcpServers": {
    "k8s-telemetry": {
      "command": "kubectl",
      "args": ["exec", "-i", "-n", "monitoring", "deploy/k8s-telemetry-mcp", "--", "k8s-telemetry-mcp"]
    }
  }
}

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "k8s-telemetry": {
      "command": "kubectl",
      "args": ["exec", "-i", "-n", "monitoring", "deploy/k8s-telemetry-mcp", "--", "k8s-telemetry-mcp"]
    }
  }
}

Restart your AI assistant after saving. It will discover all 12 tools automatically.

Step 5 — Start Asking Questions

• "Why did the payment pod crash in the last hour?"
• "Show me CPU and memory usage for the api-gateway service"
• "Is the checkout service meeting its SLO this week?"
• "Which namespace is consuming the most resources?"