mcp-k8s-eye

mcp-k8s-eye is a tool that can manage kubernetes cluster and analyze workload status.

• Connect to a Kubernetes cluster
• Generic Kubernetes Resources management capabilities
  • Support all navtie resources: Pod, Deployment, Service, StatefulSet, Ingress...
  • Support CustomResourceDefinition resources
  • Operations include: list, get, create, update, delete
• Pod management capabilities (exec, logs)
• Deployment management capabilities (scale)
• Describe Kubernetes resources
• Explain Kubernetes resources

• Pod diagnostics (analyze pod status, container status, pod resource utilization)
• Service diagnostics (analyze service selector configuration, not ready endpoints, events)
• Deployment diagnostics (analyze available replicas)
• StatefulSet diagnostics (analyze statefulset service if exists, pvc if exists, available replicas)
• CronJob diagnostics (analyze cronjob schedule, starting deadline, last schedule time)
• Ingress diagnostics (analyze ingress class configuration, related services, tls secrets)
• NetworkPolicy diagnostics (analyze networkpolicy configuration, affected pods)
• ValidatingWebhook diagnostics (analyze webhook configuration, referenced services and pods)
• MutatingWebhook diagnostics (analyze webhook configuration, referenced services and pods)
• Node diagnostics (analyze node conditions)
• Cluster diagnostics and troubleshooting

• Pod, Deployment, ReplicaSet, StatefulSet, DaemonSet workload resource usage (cpu, memory)
• Node capacity, utilization (cpu, memory)
• Cluster capacity, utilization (cpu, memory)

• Multiple transport protocols support (Stdio, SSE)
• Support multiple AI Clients

• resource_get: Get detailed resource information about a specific resource in a namespace
• resource_list: List detailed resource information about all resources in a namespace
• resource_create_or_update: Create or update a resource in a namespace
• resource_delete: Delete a resource in a namespace
• resource_describe: Describe a resource detailed information in a namespace
• deployment_scale: Scale a deployment in a namespace
• pod_exec: Execute a command in a pod in a namespace`
• pod_logs: Get logs from a pod in a namespace

• pod_analyze: Diagnose all pods in a namespace
• deployment_analyze: Diagnose all deployments in a namespace
• statefulset_analyze: Diagnose all statefulsets in a namespace
• service_analyze: Diagnose all services in a namespace
• cronjob_analyze: Diagnose all cronjobs in a namespace
• ingress_analyze: Diagnose all ingresses in a namespace
• networkpolicy_analyze: Diagnose all networkpolicies in a namespace
• validatingwebhook_analyze: Diagnose all validatingwebhooks
• mutatingwebhook_analyze: Diagnose all mutatingwebhooks
• node_analyze: Diagnose all nodes in cluster

• workload_resource_usage: Get pod/deployment/replicaset/statefulset resource usage in a namepace (cpu, memory)

• Go 1.23 or higher
• kubectl configured

# clone the repository
git clone https://github.com/wenhuwang/mcp-k8s-eye.git
cd mcp-k8s-eye

# build the binary
go build -o mcp-k8s-eye

{
  "mcpServers": {
    "k8s eye": {
      "command": "YOUR mcp-k8s-eye PATH",
      "env": {
        "HOME": "USER HOME DIR"
      },
    }
  }
}

env.HOME is used to set the HOME directory for kubeconfig file.

1. start your mcp sse server
2. config your mcp server

{
  "mcpServers": {
    "k8s eye": {
      "url": "http://localhost:8080/sse",
      "env": {}
    }
  }
}