ADK-MCP-Kubernetes

Cyber-Cloud-Pro/ADK-MCP-Kubernetes

3.1

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

The Modal Context Protocol (MCP) server is a structured messaging protocol designed to facilitate dynamic, context-driven interactions between agents and services, particularly within Kubernetes environments.

☸️Kubernetes MCP Integration with Google ADK (Agent Development Kit) !

This project implements a lightweight intelligent control plane for Kubernetes using the Google Agent Development Kit (ADK) and a custom Modal Context Protocol (MCP) server, both developed in Python using FastAPI. These services run directly on the host machine (node) and interface with the Kubernetes cluster via kubectl.

🔑Key Components

Modal Context Protocol (MCP):

A structured messaging protocol that enables dynamic, context-driven interactions between agents and services.

Kubernetes (Minikube):

A local container orchestration platform to host MCP services, simulate scaling, and manage deployments.

Docker:

Used to run the minikube cluster

Google ADK (Agent Development Kit):

A framework or toolkit used to build software agents that interact with MCP services, enabling session/context-based communication.

kubectl:

For managing Kubernetes resources (pods, deployments, services).

Node.js/NPM:

Optional — for interacting with MCP endpoints.

🛠 Features

The setup script ( req.sh ) performs the following tasks:

  • Updates the system packages
  • Installs Docker and enables it as a system service
  • Adds the current user to the docker group
  • Installs Node.js (v18), npm, and npx
  • Installs Minikube
  • Installs kubectl via Snap

Run Script

The run script (run.sh ) performs the following tasks:

  • Starts the Minikube
  • Creates Python Virtual Env
  • initiates the Env
  • Installs the Req.txt
  • Runs the Programe

📦 Prerequisites

  • Ubuntu 22.04 or 24.04 LTS
  • A user with sudo privileges
  • Internet access
  • Fresh session (log out & back in after running the script to apply Docker group changes)

📁 Files

  • req.sh — Main shell script to set up Docker, Node.js, Minikube, and kubectl

  • run.sh — Shell script for running the adk programme in virtual env

  • agent.py — Main programe code for running the MCP

  • .env — Declare the Env Variables (Add API Keys in this file)

🚀 How to Use

1. Clone the Repository

git clone https://github.com/Cyber-Cloud-Pro/ADK-MCP-Kubernetes.git
cd ADK-MCP-Kubernetes

2. Make the Shell script Executale

chmod +x run.sh
chmod +x req.sh

3. Run the req.sh

./req.sh

4. Re Login to the terminal

If using AWS EC2 refreshing the browser also works

5. Add Gemini API in .env file

Get your Google Gemini API Key

6. Run the run.sh

./run.sh

7. Explore the UI at Public-IP:8000 or LocolHost:8000

🧠 Concept

  • The MCP server exposes a REST API.

  • The Agent (via FastAPI) sends contextual requests to MCP.

  • The MCP server, in response, executes kubectl commands based on the context/intents (e.g., get pods, scale deployment, etc.).

  • The Kubernetes cluster runs separately, e.g., on Minikube.

🧱 Logical Architecture

 [FastAPI Agent (ADK)] ──► [MCP Server API]
                                     │
                        (uses subprocess/kubectl)
                                     ▼
                         [Kubernetes Cluster (via kubectl)]

🚀 Use Cases for This Setup

  • MCP interprets high-level agent intents and translates them to kubectl actions.

  • Agents can simulate automated decisions (e.g., scaling, failure handling, audit reporting).

  • Great for demoing AI-driven or rule-based orchestration flows.