Kubernetes Management & GKE Monitoring MCP Server

Created for GKE Turns 10 Hackathon #GKETurns10 #GKEHackathon

Project Overview

The GKE Intelligent Monitoring System is an advanced monitoring and management solution that combines the power of Google Kubernetes Engine (GKE) with AI-driven insights using Google's AI Development Kit (ADK) and Model Context Protocol (MCP). This system provides intelligent monitoring, automated troubleshooting, and proactive management capabilities for GKE clusters.

Features

Intelligent Cluster Monitoring
- Real-time monitoring of GKE cluster health
- Pod lifecycle management and status tracking
- Service and deployment monitoring
- Resource utilization insights
AI-Powered Troubleshooting
- Automated problem detection and diagnosis
- Intelligent remediation suggestions
- Predictive analytics for potential issues
- Network connectivity analysis
Advanced Management Capabilities
- Dynamic scaling of deployments
- YAML manifest management
- Pod execution and log analysis
- Security context management
Integration Capabilities
- Seamless integration with GKE clusters
- Support for Prometheus metrics
- Cloud Monitoring integration
- Custom tooling support through MCP
Core tools
- Discoverable tool catalog via GET /tools on the MCP server.
- Tool proxy endpoints POST /tool/<name> accepting JSON { "args": [...], "kwargs": {...} } and returning { "ok": true, "result": ... } or { "ok": false, "error": "..." }.
Core tools included:
get_cluster_info - Get basic cluster information, node status, and health
list_pods - List pods with status, resource usage, and readiness
get_pod_logs - Retrieve pod logs for troubleshooting
describe_pod - Get detailed pod information and events
get_service_status - Check service endpoints and networking
get_deployment_status - Monitor deployment health and replica status
delete_resource - Delete a Kubernetes resource (deployment, service, pod, etc.).
suggest_troubleshooting - AI-powered troubleshooting recommendations
automate remediation - Image pull remediation analysis for pod
get_gke_cluster_metrics - GKE-specific performance metrics
scale deployment - Scale a deployment to a specific number of replicas
exec pod command - Execute a command inside a pod container.
network_connectivity_test - Test network connectivity and DNS resolution

Technologies Used

Google Kubernetes Engine (GKE)
Model Context Protocol (MCP) server via mcp.server.fastmcp
Google ADK (google.adk.agents.LlmAgent) for the conversational agent
Python 3.11, kubernetes Python client, httpx, requests
Docker + Artifact Registry (or GCR) and Cloud Build for CI
kubectl manifests and RBAC for in-cluster deployment

Data Sources & External Services

Kubernetes API (in-cluster via ServiceAccount or kubeconfig)
Google Cloud Project metadata when configured (GCP_PROJECT_ID, cluster name/zone)
Optional metrics-server for resource metrics (node/pod top)

Architecture

The system consists of three main components:

ADK Agent - Handles AI-driven interactions and decision-making
MCP Server - Provides tooling and Kubernetes management capabilities
GKE Integration - Direct interface with GKE clusters and resources

graph TB
    subgraph "User Interaction"
        UI[User/Client]
    end

    subgraph "ADK Layer"
        ADK[ADK Agent]
        FASTAPI[FastAPI Server]
    end

    subgraph "MCP Layer"
        MCP[MCP Server]
        Tools[K8s Tools]
        Health[Health Check]
        Monitor[Monitoring Tools]
    end

    subgraph "GKE Cluster"
        API[Kubernetes API]
        Pods[Pods]
        Services[Services]
        Deploy[Deployments]
    end

    subgraph "Google Cloud"
        Monitoring[Cloud Monitoring]
        Logging[Cloud Logging]
    end

    UI --> FASTAPI
    FASTAPI --> ADK
    ADK --> MCP
    MCP --> Tools
    MCP --> Health
    MCP --> Monitor
    Tools --> API
    Health --> API
    Monitor --> API
    Monitor --> Monitoring
    API --> Pods
    API --> Services
    API --> Deploy
    Pods --> Logging
    Services --> Logging
    Deploy --> Logging

    classDef gcp fill:#4285F4,stroke:#4285F4,color:white;
    classDef k8s fill:#326CE5,stroke:#326CE5,color:white;
    classDef adk fill:#34A853,stroke:#34A853,color:white;
    classDef mcp fill:#EA4335,stroke:#EA4335,color:white;

    class Monitoring,Logging gcp;
    class API,Pods,Services,Deploy k8s;
    class ADK,FASTAPI adk;
    class MCP,Tools,Health,Monitor mcp;

How to Build & Deploy

Build images (Cloud Build or docker build/push).
Apply RBAC and Deployment manifests: kubectl apply -f k8s-manifests/k8s-mcp-rbac.yaml and k8s-manifests/k8s-mcp-deployment.yaml.
Deploy ADK agent: kubectl apply -f deployment.yaml.
Monitor logs: kubectl logs -f deployment/k8s-mcp-server and kubectl logs -f deployment/adk-agent.

Configuration

Environment Variables

Key environment variables:

GCP_PROJECT_ID - Your Google Cloud project ID
GKE_CLUSTER_NAME - Target GKE cluster name
GKE_ZONE - GKE cluster zone/region
MCP_SERVICE_URL - MCP server endpoint

AI Model Configuration

The project uses Google's ADK LlmAgent which internally uses the Gemini 2.0 Flash model through Vertex AI. Important points about the AI configuration:

Vertex AI Authentication: The LlmAgent requires a Vertex AI API key which is NOT set through environment variables for security reasons. Instead, it should be configured through Google Cloud's secret management system.

Deployment

The system uses Kubernetes manifests for deployment, including:

Deployment configurations
RBAC settings
Service accounts
Network policies

Findings and Learnings

During the development of this project for the GKE Turns 10 Hackathon, several key insights were gained:

GKE's robust API and integration capabilities make it an ideal platform for building intelligent monitoring solutions
Combining ADK with Kubernetes operations enables sophisticated automation and decision-making
MCP provides a flexible framework for extending monitoring capabilities
Real-time monitoring with AI-driven insights can significantly improve cluster management

Future Enhancements

Enhanced predictive analytics for resource scaling
Machine learning models for anomaly detection
Extended automation capabilities
Integration with additional Google Cloud services

Contribution & Hackathon Note

This project was created and adapted for entry in the GKE Turns 10 Hackathon. Contributions and improvements are welcome. #GKEHackathon

License

MIT License

w3sqr/k8s-mcp-and-adk-agent