Claude Certified Developer (CCDV-F) Exam Prep & AI Agents

Prepare for the Claude Certified Developer – Foundations (CCDV-F) exam while building a real-world Claude AI agent from scratch.

This is a hands-on, build-along course that combines CCDV-F exam preparation with practical Claude agent development. Instead of learning isolated concepts or memorizing exam topics, you will build one continuously evolving Repo & Pull Request Review Agent and add new capabilities to it throughout the course.

The course takes you from the raw Claude Messages API all the way to the Claude Agent SDK, custom tools, MCP, hooks and security, subagents, memory and sessions, structured outputs, and Claude Agent Skills. Each section builds on the previous one, giving you both the practical implementation experience and the conceptual understanding needed for the CCDV-F exam.

What You Will Build

You will build a complete Repo & Pull Request Review Agent that evolves throughout the course.

You start by building a tool-using agent directly with the raw Claude Messages API. You will implement the agentic loop yourself, define tools using JSON, handle tool calls and tool results, understand message ordering, work with stop_reason, and see why the raw Messages API is stateless.

You then rebuild the same agent using the Claude Agent SDK. You will learn how query() works, understand the SDK's message types and ClaudeAgentOptions, explore built-in tools, and compare the SDK implementation with the hand-built API implementation.

From there, the agent grows layer by layer.

Custom Tools & MCP

Build a real custom tool and connect it to your agent using the Agent SDK's MCP capabilities. You will learn the core components of custom tools, including the tool name, description, input schema, and handler.

Hooks, Guardrails & Security

Add practical security controls to the agent using Claude Agent SDK hooks. You will work with dangerous-command blocking, tool-output handling, evidence requirements, and prompt-injection defense, including indirect prompt injection from untrusted content.

Subagents & Multi-Agent Orchestration

Transform the agent into a coordinator with specialized subagents. You will learn about isolated context, specialized instructions, tool restrictions, parallel work, and the architectural trade-offs involved in multi-agent systems.

Agent Memory, Sessions & Context

Learn the different concepts commonly described as agent memory, including CLAUDE, the memory tool, sessions, resume, fork, and context compaction. You will make the agent capable of continuing work across sessions.

Structured Outputs

Build validated tools and structured workflows that reduce unreliable free-text output. You will work with schemas, validation, structured findings, and reliable final review generation.

Claude Agent Skills

Create reusable SKILL files and chain multiple skills together so the agent can transform a completed review into a summary and then into a release note.

MCP in Depth

Connect your agent to an external MCP server and build your own standalone MCP server using the MCP Python SDK. You will understand how MCP extends agent capabilities and how standalone MCP servers can be used across MCP-compatible clients.

Claude Agent Architecture

Before starting the hands-on build, you will also learn the three major ways of building Claude agents:

• Raw Messages API — you control the agent loop.

• Claude Agent SDK — the SDK manages the agent loop.

• Managed Agents — Anthropic manages the runtime.

You will understand the differences between these approaches, what each layer provides, and how to think about choosing the right architecture.

What You Will Learn

• Build a tool-using Claude agent from scratch using the raw Messages API.

• Understand the complete agentic loop and Claude API mechanics.

• Work with tool definitions, tool calls, tool results, messages, and stop reasons.

• Build and configure agents with the Claude Agent SDK.

• Create and integrate custom tools.

• Understand and build MCP integrations and standalone MCP servers.

• Implement hooks, guardrails, and security controls.

• Defend agents against direct and indirect prompt injection.

• Design and orchestrate specialized subagents.

• Work with agent memory, sessions, resume, fork, and context compaction.

• Build structured, validated agent outputs.

• Create and chain reusable Claude Agent Skills.

• Understand the differences between the raw API, Agent SDK, and Managed Agents approaches.

• Apply these concepts to a complete real-world Repo & Pull Request Review Agent.

CCDV-F Exam Preparation

The course is designed around the published Claude Certified Developer – Foundations (CCDV-F) exam blueprint and includes exam-focused takeaways throughout the build.

Topics covered across the course include:

• Agent Architecture

• Agent Construction with Claude

• Agent Patterns and Frameworks

• Claude API Mechanics

• Claude Application Design

• Tools and MCPs

• Claude Hooks

• Guardrails and Safe Deployment

• AI Application Security

• Agentic Customization

• Claude Code Operation

Each relevant section connects the practical implementation back to the corresponding exam concepts, so you learn the material by building and applying it rather than relying only on memorization.

Who This Course Is For

This course is designed for developers, software engineers, AI engineers, and technical professionals who want to prepare for the CCDV-F exam while gaining practical experience building Claude AI agents.

It is also suitable for developers who are not taking the certification exam but want a structured, hands-on path from the raw Claude API to advanced agent capabilities such as tools, MCP, hooks, security, subagents, memory, structured outputs, and Skills.

Prerequisites

Basic programming experience is recommended, preferably with Python.

You should be comfortable with basic programming concepts, functions, files, and working in a terminal. Basic knowledge of APIs and JSON is helpful, and familiarity with Git and GitHub is useful for following the project.

No previous experience with the Claude Messages API, Claude Agent SDK, or MCP is required. The course introduces these technologies from the ground up and builds them step by step.

You do not need to be an AI or machine-learning expert. The focus is on practical Claude agent development and the concepts required for the CCDV-F certification.

By the end of the course, you will have built a complete Claude-powered Repo & Pull Request Review Agent and developed practical experience with the core technologies and concepts covered throughout the CCDV-F-focused curriculum.