Mcp K8s Eye

What is Mcp K8s Eye?

Mcp K8s Eye is a tool that can manage a Kubernetes cluster and analyze workload status. It supports core Kubernetes operations, diagnostics for common resources, and monitoring of workload resource usage. It is intended for developers and operators who interact with Kubernetes through MCP-compatible AI clients.

How to use Mcp K8s Eye?

Clone the repository, build the binary with Go 1.23 or higher, and configure it as an MCP server in either Stdio mode (pointing to the binary and setting HOME for kubeconfig) or SSE mode (starting the SSE server and providing the URL). Use the provided tools such as resource_get, deployment_scale, pod_exec, pod_analyze, and workload_resource_usage via your AI client.

Key features of Mcp K8s Eye

• Manage all native Kubernetes resources and CustomResourceDefinitions
• Perform create, read, update, delete, and describe operations
• Execute commands in pods and retrieve pod logs
• Scale deployments
• Diagnose pods, deployments, statefulsets, services, cronjobs, ingresses, network policies, webhooks, and nodes
• Monitor workload resource usage (CPU, memory) for pods, deployments, replicasets, statefulsets, and daemonsets
• Support both Stdio and SSE transport protocols

Use cases of Mcp K8s Eye

• Quickly list and inspect resources across namespaces via an AI assistant
• Diagnose why a pod or deployment is unhealthy and get configuration insights
• Scale a deployment on demand without manually running kubectl
• Monitor CPU and memory usage of workloads to identify resource bottlenecks
• Validate ingress, network policy, and webhook configurations

FAQ from Mcp K8s Eye

What kinds of Kubernetes resources can I manage?

You can manage all native resources (Pod, Deployment, Service, StatefulSet, Ingress, etc.) and any CustomResourceDefinition. Operations include list, get, create, update, and delete.

What are the system requirements?

You need Go 1.23 or higher to build the binary, and a configured kubectl (with a valid kubeconfig) for the tool to access your cluster.

Where does my cluster data reside?

All data remains in your Kubernetes cluster. The tool reads cluster state via your kubeconfig and does not store data externally.

What transport protocols are supported?

Mcp K8s Eye supports both Stdio and SSE (Server-Sent Events) modes for connecting to MCP-compatible clients.

Are there any current limitations in monitoring?

Monitoring currently covers workload resource usage (CPU, memory) for pods, deployments, replicasets, statefulsets, and daemonsets. Node and cluster capacity/utilization metrics are not yet implemented.