Prerequisites — Installation Cheat Sheet

A running list of the tools used across EDUeki’s DevOps, Kubernetes, and AI-agent workshops. Install once, reuse for every course — this isn’t specific to any single workshop. New workshops that need a new tool get it added here rather than repeating setup instructions per-project.

Each entry has what it’s for, the install command for macOS and Windows, and a command to verify it worked.

Package managers (install these first)

Both platforms below use a package manager to keep the one-liners short. If you don’t have these yet:

• macOS — Homebrew: run the installer at brew.sh, or:

  /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
  

• Windows — winget: built into Windows 10/11 (comes with “App Installer”). If missing, get it from the Microsoft Store or see Microsoft’s winget docs.

If you’d rather not use a package manager, every section below also links the tool’s official downloads page.

Core dev tools

Git

Version control — cloning workshop repos, tracking your own changes.

• macOS: brew install git (or download from git-scm.com)
• Windows: winget install --id Git.Git -e (or git-scm.com)
• Verify: git --version

Node.js & npm

Runs npx, which is how MCP Inspector and many MCP servers are launched with zero manual install.

• macOS: brew install node (LTS) — or download from nodejs.org
• Windows: winget install OpenJS.NodeJS.LTS — or nodejs.org
• Verify: node -v and npm -v

Python 3.11+

Runs custom MCP servers written in Python (like this workshop’s Kubernetes MCP server) and any Python-based chatbot/agent scripts.

• macOS: brew install python@3.11 — or python.org/downloads
• Windows: winget install Python.Python.3.11 — or python.org/downloads (check “Add python.exe to PATH” during install)
• Verify: python3 --version and pip3 --version (Windows: python --version)

uv

Fast Python package/environment manager from Astral — some MCP servers are launched through it directly (uv run --with mcp[cli] mcp run server.py) instead of a manually managed venv.

• macOS: brew install uv — or curl -LsSf https://astral.sh/uv/install.sh | sh (see uv install docs)
• Windows: winget install --id=astral-sh.uv -e — or powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex" (same uv install docs)
• Verify: uv --version

Kubernetes & cloud

kubectl

The Kubernetes CLI — talks to any cluster (EKS or otherwise) via your kubeconfig.

• macOS: brew install kubectl — or see kubernetes.io install docs
• Windows: winget install -e --id Kubernetes.kubectl — or kubernetes.io install docs
• Verify: kubectl version --client

AWS CLI v2

Authenticates to AWS and pulls EKS cluster credentials into your kubeconfig (aws eks update-kubeconfig).

• macOS: official .pkg installer at AWS CLI install docs (Homebrew’s awscli formula lags behind v2 releases, so prefer the official installer)
• Windows: winget install Amazon.AWSCLI — or the official .msi at the same AWS CLI install docs
• Verify: aws --version, then aws configure to set your credentials

eksctl

Creates and tears down EKS clusters with one command — used by this workshop’s infra/scripts/.

• macOS: brew tap weaveworks/tap && brew install weaveworks/tap/eksctl — or see eksctl.io/installation
• Windows: winget install eksctl — or see eksctl.io/installation
• Verify: eksctl version

Helm

Kubernetes package manager — installs charts like kube-prometheus-stack.

• macOS: brew install helm — or see helm.sh/docs/intro/install
• Windows: winget install Helm.Helm — or see helm.sh/docs/intro/install
• Verify: helm version

k9s

Terminal UI for Kubernetes — a faster way to browse pods/logs/events live than repeated kubectl get/describe calls. Handy for demos.

• macOS: brew install k9s — or see k9scli.io/topics/install
• Windows: winget install Derailed.k9s — or see k9scli.io/topics/install
• Verify: k9s version (run k9s to launch it against your current kubectl context)

Build tools

Go

Needed to build any Go-based MCP server from source — e.g. mcp-grafana (go install github.com/grafana/mcp-grafana/cmd/mcp-grafana@latest).

• macOS: brew install go — or go.dev/dl
• Windows: winget install GoLang.Go — or go.dev/dl
• Verify: go version

Containers (optional, useful generally)

Docker Desktop

Not required by every workshop, but common enough across DevOps courses to list here — local image builds, running containers outside Kubernetes.

• macOS: brew install --cask docker — or docker.com/products/docker-desktop
• Windows: winget install Docker.DockerDesktop — or docker.com/products/docker-desktop
• Verify: docker --version (Docker Desktop app must be running)

AI / MCP tooling

Claude Desktop

The chat app used to connect and test MCP servers (custom or product) as tools, without writing any client code.

• macOS / Windows: download directly from claude.ai/download — no CLI install
• Verify: open the app, Settings → Developer/Extensions should list your configured MCP servers

Claude Code CLI

Terminal-based coding agent — useful for building/debugging MCP servers themselves, not just using them.

• macOS / Windows: see claude.com/claude-code for install instructions
• Verify: claude --version

MCP Inspector

The standard local testing/debugging tool for any MCP server — think Postman, for MCP. No separate install needed; npx fetches it on first run (requires Node.js above).

• macOS / Windows: npx @modelcontextprotocol/inspector (or npm install -g @modelcontextprotocol/inspector to skip the fetch on future runs)
• Docs: modelcontextprotocol.io
• Verify: the command above should open localhost:6274 in your browser

Optional

VS Code

Not required, but the most common editor across these workshops if you don’t already have one.

• macOS: brew install --cask visual-studio-code — or code.visualstudio.com
• Windows: winget install Microsoft.VisualStudioCode — or code.visualstudio.com
• Verify: code --version