EKS Model Context Protocol (MCP) Server

EKS Model Context Protocol (MCP) Server

A lightweight, efficient server that implements the Model Context Protocol for EKS operations. This server provides a standardized interface for GenAI agents to interact with EKS clusters.

Features

• Fast response times with proper timeout handling
• Comprehensive EKS operations support
• Kubernetes resource management (pods, services, deployments, etc.)
• Multiple authentication methods for EKS clusters
• Robust error handling and logging
• Docker containerization for easy deployment
• ECS deployment support

Project Structure

• main.py: Main server implementation with FastAPI
• eks_operations.py: EKS API operations implementation
• k8s_operations.py: Kubernetes API operations implementation
• k8s_operations_sdk_v4.py: SDK-based Kubernetes operations implementation
• k8s_operations_kubectl.py: Kubectl-based Kubernetes operations with proper authentication
• k8s_auth_config.py: Authentication configuration for Kubernetes
• direct_k8s_client.py: Direct Kubernetes API client implementation
• test_*.py: Various test scripts for different implementations
• Dockerfile and Dockerfile.amd64: Docker configuration files
• docker-compose.yml: Docker Compose configuration
• requirements.txt: Python dependencies
• client/: Client implementation directory (Client README)
• ecs-deployment/: ECS deployment scripts and configuration (ECS Deployment README)

API Endpoints

• GET /health - Health check endpoint
• GET /mcp/v1/operations - List available operations
• POST /mcp/v1/query - Execute MCP operations

Supported Operations

EKS Cluster Operations

• list_clusters - List EKS clusters in a region
• describe_cluster - Get detailed information about a cluster
• list_nodegroups - List nodegroups for a cluster
• describe_nodegroup - Get detailed information about a nodegroup

Kubernetes Operations

• list_namespaces - List Kubernetes namespaces
• list_pods - List pods in a namespace
• describe_pod - Get detailed information about a pod
• get_deployments - List deployments in a namespace
• describe_deployment - Get detailed information about a deployment
• get_services - List services in a namespace
• describe_service - Get detailed information about a service
• get_pod_logs - Get logs from a pod

Authentication Methods

The server supports multiple authentication methods for EKS clusters:

1. AWS SDK Authentication: Uses the AWS SDK to authenticate with EKS and get cluster information
2. Kubectl with Generated Kubeconfig: Creates a temporary kubeconfig file and uses kubectl
3. Direct Kubernetes API Calls: Makes direct API calls to the Kubernetes API server

The current implementation uses the Kubectl method with proper authentication, which provides the most reliable results.

Getting Started

Prerequisites

• Docker
• AWS credentials (for production use)
• Python 3.8+ (for local development)

Running Locally

# Install dependencies
pip install -r requirements.txt

# Run the server
python main.py

Running with Docker

# Build the Docker image
docker build -t eks-mcp-server .

# Run the container
docker run -d -p 8000:8000 --name mcp-server eks-mcp-server

Running with Docker Compose

# Start the server
docker-compose up -d

# Stop the server
docker-compose down

Deploying to ECS

See the ECS Deployment README for detailed instructions.

# Deploy to ECS
cd ecs-deployment
./deploy.sh

# Update existing ECS deployment
./update-service.sh

Testing

Use the included test scripts to verify the server is working correctly:

# Basic tests
python test_mcp.py

# Comprehensive operation tests
python test_mcp_operations.py

# Test namespace integration
python test_mcp_namespace.py
python test_mcp_namespace_2.py

# Test kubectl integration
python test_kubectl_operations.py

# Test SDK v3 operations
python test_v3_operations.py

# Test SDK v4 operations
python test_v4_operations.py

Client Usage

The project includes a client implementation that uses Amazon Bedrock for conversational interaction with the MCP server. See the Client README for details.

# Run the client
cd client
python mcp_chat_client_v6.py

Example Requests

List Clusters

curl -X POST http://localhost:8000/mcp/v1/query \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your-api-key" \
  -d '{
    "operation": "list_clusters",
    "parameters": {
      "region": "us-east-1"
    }
  }'

List Pods

curl -X POST http://localhost:8000/mcp/v1/query \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your-api-key" \
  -d '{
    "operation": "list_pods",
    "parameters": {
      "cluster_name": "my-cluster",
      "namespace": "default",
      "region": "us-east-1"
    }
  }'

Architecture

The server is built with FastAPI and uses a combination of synchronous and asynchronous operations for improved performance. It includes:

• Timeout handling for AWS API calls
• Multiple authentication methods for EKS clusters
• Comprehensive logging
• Response time tracking
• Error handling with detailed error codes

EKS Authentication

For the server to authenticate with EKS clusters, the ECS task role needs to be added to the EKS cluster's aws-auth ConfigMap. See the ECS Deployment README for detailed instructions.

Current Deployment

The server is currently deployed to ECS and accessible at:

• Endpoint: http://3.90.45.69:8000
• Health check: http://3.90.45.69:8000/health
• Operations discovery: http://3.90.45.69:8000/mcp/v1/operations

Test Namespaces

The project includes test namespaces in the EKS cluster:

• mcp-test-namespace: Contains a deployment with 2 nginx pods
• mcp-test-namespace-2: Contains a deployment with 3 nginx pods

License

MIT