MCP Server on EKS

Name: MCP Server on EKS
Author: jbrcoleman

@jbrcoleman

リポジトリ

MCP Server on EKS について

Demo eks mcp server

#mcp-server #eks #terraform #kubernetes

基本情報

カテゴリ

その他

ランタイム

python

トランスポート

stdio

公開者

jbrcoleman

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-server-eks-jbrcoleman": {
      "command": "python",
      "args": [
        "server-enhanced.py"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is MCP Server on EKS?

This project deploys a Model Context Protocol (MCP) server on Amazon EKS using Terraform for infrastructure management. It provides natural-language and programmatic access to Kubernetes cluster operations, targeting developers and platform teams who manage AWS EKS clusters.

How to use MCP Server on EKS?

Deploy via the provided scripts in phases (infrastructure, Karpenter, app) or all at once using ./scripts/deploy.sh all. For legacy deployment without Karpenter, use make init plan apply then make deploy. Local testing is possible with python server-enhanced.py after installing dependencies. Integrate with Claude Desktop by copying the claude-desktop-config.json to the appropriate configuration path.

Key features of MCP Server on EKS

Real Kubernetes API integration with actual EKS clusters
Live cluster information with namespace counts and node status
Tools: get_cluster_status, list_pods, get_deployments
Health checks with liveness and readiness probes
Supports both in-cluster and local kubectl configurations
Claude Desktop and programmatic Python client integration

Use cases of MCP Server on EKS

Monitor real-time EKS cluster status and node health via Claude Desktop
List and inspect pods across any namespace with live status
Check deployment status and replica counts using natural language
Automate Kubernetes operations through a programmatic Python client
Test and validate MCP protocol implementations in a cloud-native environment

FAQ from MCP Server on EKS

What are the prerequisites?

AWS CLI configured, Docker installed, Terraform >= 1.0, kubectl installed, and an AWS account with an existing ECR repository.

How do I deploy with Karpenter?

Run ./scripts/deploy.sh all for a single-command deployment, or deploy in phases: ./scripts/deploy.sh infrastructure, then ./scripts/deploy.sh karpenter, then ./scripts/deploy.sh app.

How can I monitor the MCP server?

Use kubectl get service mcp-server-service -n mcp-server to access the service and kubectl logs -f deployment/mcp-server -n mcp-server to view logs.

How do I clean up resources?

For Karpenter deployments, remove resources in reverse order: kubectl delete -f k8s/, then destroy Karpenter Terraform, then the main Terraform. For legacy deployments, use make destroy (not explicitly shown but implied by cleanup section).

What does Karpenter add to the deployment?

Karpenter provides Kubernetes-native node autoscaling with a NodePool and EC2NodeClass, configuring instance types (t3.medium/large/xlarge), capacity types (spot/on-demand), and cost optimization by preferring spot instances.