wenhuwang/mcp-k8s-eye

Free

用于 Kubernetes 管理的 MCP 服务器，分析集群和应用健康状况。

FreeFree tier

Type

Open Source

Visit Website GitHub (40 stars)

About wenhuwang/mcp-k8s-eye

mcp-k8s-eye is an open-source MCP (Model Context Protocol) server for Kubernetes cluster management and diagnostics. It provides core operations on all native Kubernetes resources (Pod, Deployment, Service, StatefulSet, Ingress, etc.) and CustomResourceDefinitions, including list, get, create, update, delete, describe, and explain. It also offers specialized pod management (exec, logs) and deployment scaling. The tool features comprehensive diagnostics for pods, deployments, StatefulSets, services, CronJobs, Ingresses, NetworkPolicies, ValidatingWebhooks, MutatingWebhooks, and nodes, analyzing configurations, resource utilization, and potential issues. Monitoring capabilities cover CPU and memory usage per workload and cluster level. It supports both Stdio and SSE transport protocols and is compatible with multiple AI clients. Built in Go, it requires Go 1.23+ and a configured kubectl.

Key Features

Core Kubernetes operations: list, get, create, update, delete, describe, explain on all native resources (Pod, Deployment, Service, StatefulSet, Ingress, etc.) and CustomResourceDefinitions

Pod management: exec into containers and retrieve logs

Deployment scaling: scale deployments up or down

Diagnostics for pods, deployments, StatefulSets, services, CronJobs, Ingresses, NetworkPolicies, ValidatingWebhooks, MutatingWebhooks, and nodes

Monitoring of CPU and memory usage for pods, deployments, ReplicaSets, StatefulSets, DaemonSets, nodes, and the entire cluster

Multiple transport protocols: Stdio and SSE

Compatible with various AI clients

Pros & Cons

Pros

Comprehensive diagnostics covering many resource types including webhooks and custom resources
Supports both listing/getting and full CRUD operations (create, update, delete)
Includes monitoring of CPU and memory at workload and cluster level
Simple setup: requires only Go 1.23+ and configured kubectl
Open source with permissive license (MIT implied)

Cons

Requires manual build from source; no pre-built binary packages provided
Depends on local kubectl configuration and Kubernetes cluster access
Limited to Kubernetes-related tasks; not a general-purpose tool
Documentation is mostly in the README; no separate website or detailed guides

Best For

Managing and operating Kubernetes resources programmatically via AI clientsDiagnosing and troubleshooting issues in Kubernetes workloads and cluster componentsMonitoring resource utilization and capacity planning for Kubernetes clustersAutomating common Kubernetes operations like deployments, scaling, and pod debugging

FAQ

What Kubernetes resources does mcp-k8s-eye support?

It supports all native resources such as Pod, Deployment, Service, StatefulSet, Ingress, and also CustomResourceDefinitions (CRDs).

What transport protocols does it support?

It supports Stdio (standard I/O) and SSE (Server-Sent Events) as transport protocols.

What kind of diagnostics can it perform?

It can diagnose pods (status, containers, resource usage), deployments (available replicas), StatefulSets (service/PVC existence), services (selector config, endpoints, events), CronJobs (schedule, deadlines), Ingresses (class, related services, TLS), NetworkPolicies (configuration, affected pods), ValidatingWebhooks and MutatingWebhooks (configuration, referenced services/pods), and nodes (conditions).

How do I install mcp-k8s-eye?

Clone the repository, ensure Go 1.23+ and kubectl are configured, then build the binary from source using 'go build'.