Agentic-Kubernetes-Management

Tarique-B-DevOps/Agentic-Kubernetes-Management

3.2

If you are the rightful owner of Agentic-Kubernetes-Management and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

The Model Context Protocol (MCP) server facilitates the integration of Kubernetes tools for natural language management through a structured communication protocol.

Tools
3
Resources
0
Prompts
0

Agentic Kubernetes Management

Strands AI agent using MCP servers to expose Kubernetes tools (pods, namespaces, logs) for natural language cluster management.

Python 3.12+ License: MIT Kubernetes

🌟 Features

  • Natural Language Interface - Manage Kubernetes clusters using conversational AI
  • MCP Protocol - Leverages Model Context Protocol for extensible tool integration
  • Strands Framework - Built on Strands AI agent framework for robust agentic behavior
  • AWS Bedrock Integration - Powered by Claude models via Amazon Bedrock
  • Real-time Operations - List pods, inspect namespaces, and retrieve logs instantly
  • Context-Aware - Maintains conversation history for intelligent follow-up queries
  • Production Ready - Error handling, graceful shutdown, and comprehensive logging

🏗️ Architecture

┌─────────────────┐
│   User Input    │
│ (Natural Lang)  │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Strands Agent  │
│  (Claude Model) │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│   MCP Server    │
│ (Tool Provider) │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Kubernetes API  │
│                 │
└─────────────────┘

🚀 Quick Start

Prerequisites

  • Python 3.12 or higher
  • Kubernetes cluster access (kubeconfig configured)
  • AWS credentials configured for Bedrock access
  • Anthropic model enabled

Installation

  1. Clone the repository
git clone https://github.com/Tarique-B-DevOps/Agentic-Kubernetes-Management.git
cd Agentic-Kubernetes-Management
  1. Install dependencies
pip install -r requirements.txt
  1. Set up environment variables
export BEDROCK_MODEL_REGION="us-east-1"
export BEDROCK_MODEL_ID="apac.anthropic.claude-3-5-sonnet-20241022-v2:0"
  1. Verify Kubernetes access
kubectl cluster-info
  1. Run the agent
python agent.py

💬 Usage Examples

Basic Queries

🔍 Your query: Show me all namespaces
📊 Response: Found 5 namespaces:
- default (Active)
- kube-system (Active)
- kube-public (Active)
- monitoring (Active)
- production (Active)

🔍 Your query: List pods in the production namespace
📊 Response: Found 12 pods in namespace 'production':
- nginx-deployment-7d8f9c5b6-abc12 (Running)
- redis-master-0 (Running)
- api-server-v2-5f9d8c7b-xyz34 (Running)
...

🔍 Your query: Get logs from api-server-v2-5f9d8c7b-xyz34
📊 Response: Logs from pod 'api-server-v2-5f9d8c7b-xyz34' (container: 'api', namespace: 'production'):

2024-01-15 10:23:45 INFO Starting API server...
2024-01-15 10:23:46 INFO Connected to database
...

Advanced Queries

🔍 Your query: Are there any failing pods in kube-system?
🔍 Your query: Show me the last 50 lines of logs from the nginx pod
🔍 Your query: Which pods are running on node worker-01?

🛠️ Available Tools

ToolDescriptionParameters
get-podsList pods in a namespacenamespace (default: "default")
get-namespacesList all namespacesNone
check-logsRetrieve pod logspod_name, namespace, container, tail_lines

📁 Project Structure

Agentic-Kubernetes-Management/
├── k8s_agent.py           # Main Strands agent with interactive CLI
├── k8s_mcp_server.py      # MCP server exposing Kubernetes tools
├── requirements.txt        # Python dependencies
├── README.md              # Project documentation
└── .env.example           # Environment variables template

🔧 Configuration

Environment Variables

VariableDescriptionDefault
BEDROCK_MODEL_REGIONAWS region for Bedrockus-east-1
BEDROCK_MODEL_IDClaude model identifierapac.anthropic.claude-3-5-sonnet-20241022-v2:0

Kubernetes Configuration

Ensure your ~/.kube/config is properly configured:

kubectl config view
kubectl config current-context

🧩 How It Works

1. MCP Server (k8s_mcp_server.py)

The MCP server exposes Kubernetes operations as callable tools using the FastMCP framework:

  • get-pods - Queries Kubernetes API for pod information
  • get-namespaces - Lists all namespaces in the cluster
  • check-logs - Retrieves container logs from pods

2. Strands Agent (agent.py)

The Strands agent:

  • Connects to the MCP server via stdio transport
  • Loads available tools dynamically
  • Processes natural language queries using Claude
  • Maintains conversation context across multiple queries
  • Invokes appropriate tools based on user intent

3. Model Context Protocol (MCP)

MCP enables:

  • Tool Discovery - Agent discovers available Kubernetes tools
  • Structured Communication - Standardized request/response format
  • Extensibility - Easy to add new tools without changing agent code

🚧 Roadmap

  • context settings
  • Add deployment management tools
  • Support for ConfigMaps and Secrets
  • Node resource monitoring
  • Service and Ingress management
  • Multi-cluster support
  • Prometheus metrics integration
  • Slack/Discord bot integration
  • Web UI dashboard

Acknowledgments