Model Context Protocol moderate risk

K8s Eye

mcp-k8s-eye provides comprehensive Kubernetes cluster management and diagnostics, including resource operations, analysis, and monitoring via multiple transport protocols.

Connections & Capabilities

Connects To

GitHubKubernetes

Capabilities

readwriteexecadmin

Quickstart

Config

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

Exposed MCP Tools (19)

safe

resource_get

Retrieves detailed information about a specific Kubernetes resource.

Read-only operation, no modification of resources.

safe

resource_list

Lists detailed information about all Kubernetes resources in a namespace.

Read-only operation, no modification of resources.

high

resource_create_or_update

Creates or updates a Kubernetes resource in a specified namespace.

Modifies cluster state, potentially causing disruption or security issues.

critical

resource_delete

Deletes a Kubernetes resource from a specified namespace.

Destructive operation that can lead to data loss or service disruption.

safe

resource_describe

Provides detailed information about a Kubernetes resource.

Read-only operation, no modification of resources.

moderate

deployment_scale

Scales a Kubernetes deployment in a specified namespace.

Modifies the number of replicas, potentially impacting application availability.

critical

pod_exec

Executes a command within a pod in a specified namespace.

Allows arbitrary command execution within a container, posing a significant security risk.

safe

pod_logs

Retrieves logs from a pod in a specified namespace.

Read-only operation, no modification of resources.

safe

pod_analyze

Diagnoses all pods in a namespace, identifying potential issues.

Read-only analysis of pod status.

safe

deployment_analyze

Diagnoses all deployments in a namespace, identifying potential issues.

Read-only analysis of deployment status.

safe

statefulset_analyze

Diagnoses all statefulsets in a namespace, identifying potential issues.

Read-only analysis of statefulset status.

safe

service_analyze

Diagnoses all services in a namespace, identifying potential issues.

Read-only analysis of service configuration and endpoints.

safe

cronjob_analyze

Diagnoses all cronjobs in a namespace, identifying potential issues.

Read-only analysis of cronjob schedule and status.

safe

ingress_analyze

Diagnoses all ingresses in a namespace, identifying potential issues.

Read-only analysis of ingress configuration and related services.

safe

networkpolicy_analyze

Diagnoses all networkpolicies in a namespace, identifying potential issues.

Read-only analysis of networkpolicy configuration and affected pods.

safe

validatingwebhook_analyze

Diagnoses all validating webhooks, identifying potential issues.

Read-only analysis of validating webhook configurations.

safe

mutatingwebhook_analyze

Diagnoses all mutating webhooks, identifying potential issues.

Read-only analysis of mutating webhook configurations.

safe

node_analyze

Diagnoses all nodes in the cluster, identifying potential issues.

Read-only analysis of node conditions.

safe

workload_resource_usage

Gets pod/deployment/replicaset/statefulset resource usage in a namespace (cpu, memory).

Read-only operation, no modification of resources.

Safety Assessment

mcp-k8s-eye offers powerful Kubernetes management capabilities, but the ability to create, update, delete resources and execute commands in pods introduces risk. It is safe to use for monitoring and diagnostics in read-only mode. Exercise caution when using write operations, especially in production environments.

Provides diagnostics tools to analyze Kubernetes resources.
Offers monitoring tools for workload resource usage.
Supports multiple transport protocols (Stdio, SSE).
Supports multiple AI Clients.

Includes tools for creating, updating, and deleting Kubernetes resources.
Allows pod execution, potentially leading to security risks if not properly controlled.
Requires kubectl configuration, which could expose credentials if misconfigured.
No built-in sandboxing for resource operations.