🎯 Kubernetes AI Management System

AI-Powered Kubernetes Management (MCP + Agent)

    ⎈ K8s AI Management
    ├── 🤖 MCP Server
    ├── 🔍 K8s Tools
    └── 🚀 Agent mode with Rest API

✨ Overview

This project combines the power of AI with Kubernetes management. Users can perform real-time diagnostics, resource monitoring, and smart log analysis. It simplifies Kubernetes management through conversational AI, providing a modern alternative.

💡 Just ask questions naturally - no need to memorize commands!

🏗️ Project Structure

The project is organized into the following modules:

• agent: Agent mode backed by Rest API to analyze the cluster using natural language
• mcp-server: MCP server backed by tools which can be integrated with MCP host (like Claude desktop) to provide a full experience
• tools: Kubernetes tools for cluster analysis/management (used by both agent and mcp-server)

🎁 Features

This AI-powered system understands natural language queries about your Kubernetes cluster. Here are some of the capabilities provided by the system which can be queried using natural language:

🏥 Cluster Health and Diagnostics

• "What's the status of my cluster?"
• "Show me all pods in the default namespace"
• "Are there any failing pods? in default namespace"
• "What's using the most resources in my cluster?"
• "Give me a complete health check of the cluster"
• "Are there any nodes not in Ready state?"
• "Show me pods in default namespace that have been running for more than 7 days"
• "Identify any pods running in default namespace with high restart counts"

🌐 Network Analysis

• "Show me the logs for the payment service"
• "List all ingresses in the cluster"
• "Show me all services and their endpoints"
• "Check if my service 'api-gateway' has any endpoints"
• "Show me all exposed services with external IPs"

💾 Storage Management

• "List all persistent volumes in the cluster"
• "Show me storage claims that are unbound"
• "What storage classes are available in the cluster?"
• "Which pods are using persistent storage?"
• "Are there any storage volumes nearing capacity?"

⏱️ Job and CronJob Analysis

• "List all running jobs in the batch namespace"
• "Show me failed jobs from the last 24 hours"
• "What CronJobs are scheduled to run in the next hour?"
• "Show me the execution history of the 'backup' job"

⎈ Helm Release Management

• "List all Helm releases"
• "Upgrade the MongoDB chart to version 12.1.0"
• "What values are configured for my Prometheus release?"
• "Rollback the failed Elasticsearch release"
• "Show me the revision history for my Prometheus release"
• "Compare values between different Helm releases"
• "Check for outdated Helm charts in my cluster"
• "What are the dependencies for my Elasticsearch chart?"

Note: The system uses AI to analyze patterns in logs, events, and resource usage to provide intelligent diagnostics and recommendations.

🛠️ Prerequisites

Requirement	Version
☕ JDK	17 or later
🧰 Maven	3.8 or later
⎈ Minikube/Any Kubernetes cluster	Configured ~/.kube/config

Note: The system uses the kubeconfig file from ~/.kube/config, so make sure it is properly configured.

---

🏗️ 1. Project Build

# Build all modules
mvn clean package

# Run the MCP server
java -jar mcp-server/target/mcp-server-1.0-SNAPSHOT.jar

# Alternatively, run the agent directly
java -jar agent/target/agent-*-fat.jar

🛠️ 2. Minikube setup

Install minikube and create a nginx deployment:

# Install minikube
brew install minikube

# Start minikube
minikube start

# Make sure kubeconfig is set
kubectl config use-context minikube

# Deploy nginx
kubectl create deployment nginx --image=nginx:latest

# Check whether nginx is running
kubectl get pods

Note: You should see nginx pod in the output

🛠️ 3. Testing project

🤝 3.1 MCP Server integration with Claude Desktop

Refer to mcp-server/README.md for instructions on how to integrate with Claude Desktop

3.2. Agent Mode with Rest API

Refer to agent/README.md for instructions on how to run the agent

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.