Claude Code Bootcamp: Hooks, MCP & Agentic AI Workflows

Stop using AI like a chatbot. Start engineering with it. The way we write software is changing — not slowly, not gradually, but fundamentally. Today, AI can write code. Fix bugs. Review pull requests. Open GitHub issues. Automate entire workflows.

But here's the truth most courses won't tell you:

Most developers are still using AI like a chatbot.

They type a prompt. They copy the answer. They move on. And they wonder why their AI feels powerful in demos — and useless in their actual codebase.

This course flips that completely.

By the end of this bootcamp, you won't just use Claude Code. You'll orchestrate it. You won't be writing prompts. You'll be engineering systems that AI runs for you — hands-free, around the clock, on real production codebases.

What You'll Learn — Section by Section

Section 1 — Mastering Claude Code: Setup, Modes, Permissions & Customization

You'll start with the foundation — installing Claude Code, configuring it for your environment, and understanding the four core modes that determine how Claude behaves: Default, Auto-Accept, Plan, and YOLO.

You'll master:

• Installing Claude Code on macOS, Linux, and Windows

• Choosing between the terminal experience and native VS Code integration

• Working with the CLI, slash commands, and special notations like @file and #memory

• Understanding when to use which mode — and the trade-offs between safety and speed

• Building your first CLAUDE md — the file that trains Claude on your project once and forever

• Locking down permissions with three guards: Allow, Ask, Deny

• Mastering the settings.json file and /config command

• Configuring Claude Code with CLI arguments for power users

• Teaching Claude how to talk to you with custom output styles

By the end of this section, Claude Code will know your project, your preferences, and your boundaries — without you typing a single instruction twice.

Section 2 — Never Lose Your Work: Mastering Context, Sessions & Checkpoints

Every developer working with AI hits the same wall — the context window fills up, and suddenly you've lost two hours of work. This section makes sure that never happens to you again.

You'll learn:

• The invisible fuel tank — how Claude Code's context window actually works

• When to use /compact vs /clear — and the critical difference between them

• Full session management — saving, naming, resuming, forking

• Resume, fork, and continue any session like a Git branch

• Checkpoints — Ctrl+Z on steroids, your safety net before risky edits

• Recovering from broken code in seconds, not hours

• Running multiple parallel Claude Code sessions with Git worktrees

This is the section that turns Claude Code from a flaky assistant into a reliable engineering partner you can trust with long, complex tasks.

Section 3 — Claude Code + GitHub: Automating Real Developer Workflows

Imagine pushing your code, and before you even open GitHub, AI has already reviewed it, fixed bugs, and opened the pull request.

Imagine tagging @claude on a bug report — and finding a fix waiting for you when you wake up.

That's what this section delivers.

You'll learn three powerful integration patterns:

• Terminal/Local integration — let Claude Code init repos, commit, and push for you

• GitHub Actions @claude bot — autonomous PR review and issue triage in your repos

• Headless/Scripting integration — running Claude Code in CI/CD pipelines

You'll build:

• Custom commands like /commit and /commit-push-pr that turn entire workflows into one keystroke

• A fully working GitHub Actions setup for the @claude bot — line-by-line breakdown of the workflow YAML

• An issue-to-PR pipeline — bug reported, AI investigates, AI fixes, PR opened, all automatic

• Real demos on a Spring Boot job portal application so you see this in production code, not toy examples

By the end, your GitHub repository will feel less like a code host and more like a team member that ships while you sleep.

Section 4 — From Chaos to Control: Mastering Rules, Memory & Skills

Your CLAUDE md will grow. It always does. From a clean 50 lines to a 2,000-line monster Claude reads on every prompt — slow, expensive, and half irrelevant.

This section breaks that monolith into a clean, modular system.

You'll master:

• Rules — splitting your CLAUDE md into focused, modular files

• Auto Memory — Claude Code's personal notebook it writes for itself, so you don't have to

• Skills — on-demand expertise Claude loads only when needed (progressive disclosure)

• Context Fork — running skills inside isolated sub-agents to keep your main context clean

• Slash Commands — and how they're quietly evolving into the Skills system

• Building your own custom skill from scratch that Claude loads automatically when relevant

You'll see real demos of Claude building its own memory in real time, watching your CLAUDE md go from chaos to clean, and witnessing the silent evolution from Commands to Skills that's reshaping how Claude Code works in 2026.

This section alone will save you hours of context window costs and dramatically speed up every single Claude Code session.

Section 5 — Hooks: Teach Claude Code to Follow Your Rules, Not Its Instincts

You tell Claude never to touch a file. It touches it anyway.

You ask it to skip secrets. It commits an API key.

You say "don't run migrations in production." It runs them.

Hooks turn hopes into guarantees.

Hooks are scripts that fire automatically — before tool calls, after tool calls, on session start, on notifications. They're how you turn Claude Code from a polite assistant into a rule-following engineer.

You'll build:

• Your first hook — a notification system that taps you on the shoulder when Claude needs you

• A PreToolUse hook in Python that scans for secrets and blocks giant files before damage happens

• A shell script hook that physically shields critical files from any modification

• Cross-platform hooks that work on macOS, Linux, and Windows

• SessionStart hooks that pre-load context every time you open a project

• A file-length speed bump hook that warns before bloating files past your size limits

You'll understand the full hook lifecycle — every event, every payload, every return value — through a deep documentation tour built for working developers.

By the end of this section, your Claude Code setup will be bulletproof. No more accidental commits. No more leaked secrets. No more "I told it not to do that."

Section 6 — Mastering MCP: Connect, Automate & Supercharge Claude Code with Real Tools

Claude knows your code. But it doesn't know your browser. Your database. Your latest framework docs. So it hallucinates.

MCP (Model Context Protocol) is the USB-C of AI — one universal protocol that connects Claude to anything.

In this section, you'll plug Claude Code into the real world:

• Chrome DevTools MCP — watch Claude debug a live web app in your actual browser

• Playwright MCP — full browser automation, end-to-end testing, scraping, all driven by Claude

• Context7 MCP — real-time, up-to-date documentation for any framework, no more outdated answers

• A complete developer's walkthrough of the MCP specification — understand it deeply, not just superficially

• How Claude Code loads 10+ MCP servers without bloating context — the smart loading mechanics under the hood

You'll see live demos of Claude controlling a browser, debugging real applications, and pulling fresh framework docs into its context — all without you copy-pasting anything.

This is the section that makes Claude Code feel like science fiction.

Section 7 — Sub-Agents & Agent Teams: Claude Code's Multi-Agent Universe

You ask Claude to review your code. You get one opinion. From one perspective.

Now imagine spinning up three sub-agents in parallel — a security reviewer, a performance reviewer, and a style reviewer — and getting three deep, specialized reviews in the time of one.

That's the power of sub-agents.

You'll learn:

• The divide-and-conquer pattern that turns Claude Code from a single agent into a team

• How to design specialized sub-agents for code review, debugging, testing, and security auditing

• A live demo of a bug-investigator sub-agent hunting down complex defects

• How to run 3 sub-agents simultaneously reviewing the same code from different angles

• Agent teams — multiple agents collaborating toward a shared goal in parallel

• Deep dive into the official Claude Code sub-agent documentation — every parameter explained simply

Section 8 — YOLO in a Box: Safe Autonomous Claude Code with Docker Sandboxes

YOLO mode is fast. But on your bare laptop, Claude Code can reach everything: your files, your .env secrets, your API keys, your production data.

One prompt injection or one honest mistake is all it takes.

This section gives Claude Code full freedom inside a box it can’t escape.

You’ll learn:

• Why autonomous agents need isolation, not just permissions

• How Docker Sandboxes work and how to run Claude Code inside one

• Running YOLO mode safely without risking your host machine

• Playing the attacker: trying to steal files, API keys and data, and watching the sandbox block every attempt

• Applying the same sandboxing approach to any agent, from coding agents to custom agents built with Spring AI, LangChain or CrewAI

By the end of this section, you'll have a fleet of specialized AI engineers — each one an expert in its domain — working together on your codebase.

This is the future of software development. And you'll be building it with your own hands.

Prerequisites

• Basic programming knowledge in any language (Python, Java, JavaScript, etc.)

• Comfort with the command line / terminal

• A GitHub account (free is fine)

• A Pro Claude account (we'll show you how to set it up)

You do not need to be an AI expert. You do not need a machine learning background. If you can write a function and use Git, you're ready.