Claude Certified Architect (CCA-F) Complete Exam Prep Course

Claude Certified Architect (CCA-F) is Anthropic's certification for professionals who design and ship production-grade applications with Claude — covering agentic systems, the Model Context Protocol (MCP), Claude Code, and reliable prompt engineering. The exam is scenario-based: instead of testing trivia, it places you inside real production situations and asks you to make the same architectural judgment calls a working Claude solutions architect faces every day.

This course is a complete, hands-on preparation program built domain by domain against the official exam blueprint. Every concept is explained first, then applied through code walkthroughs, real configuration examples, and worked exam-style scenario questions, so you finish each section able to both explain the concept and apply it.

You will start with the fundamentals of the agentic loop and the stop reason field, then move into multi-agent coordinator architectures, subagent context passing, and session management. From there the course covers tool design and MCP integration in depth — writing tool descriptions that avoid misrouting, handling tool and agent errors correctly, and configuring MCP servers so they work consistently across a team. A full module is dedicated to Claude Code itself: the CLAUDE hierarchy, custom slash commands and skills, Plan Mode versus direct execution, and integrating Claude Code into CI/CD pipelines. The prompt engineering module covers explicit criteria prompting, few-shot examples, structured output with JSON schemas, and batch processing and multi-pass review architectures. The final module focuses on context management and reliability: why agents "forget" mid-conversation, escalation and human-handoff design, error propagation across subagents, and building human review workflows with stratified sampling and confidence-based routing.

The course closes with a dedicated exam-preparation section that works through official-style sample questions option by option, explains why each distractor is wrong, and ends with a one-page exam-day cheat sheet.

What you will learn

• The agentic loop lifecycle and how to use stop reason correctly

• How to design hub-and-spoke multi-agent coordinator systems

• How to pass structured context between subagents without losing attribution

• Building a working multi-agent system end to end with the Claude SDK

• Writing PreToolUse and PostToolUse hooks for deterministic enforcement

• Designing tool descriptions that prevent agent misrouting

• Structured error handling and tool_choice configuration

• Setting up and debugging MCP servers, including .mcp.json and .claude.json

• Claude Code's built-in tools: Grep, Glob, Read, Write, Edit, and Bash

• The CLAUDE hierarchy and project-level configuration rules

• Building custom slash commands and skills

• Choosing between Plan Mode and direct execution

• Integrating Claude Code into CI/CD pipelines correctly

• Prompt engineering techniques that reduce false positives in AI review systems

• Few-shot prompting and structured JSON output design

• Batch API usage and multi-pass, multi-instance review architecture

• Context engineering: trimming tool output, progressive summarization, and case-facts patterns

• Designing escalation rules for when an agent should hand off to a human

• Error propagation and recovery patterns across subagents

• Human review workflows using stratified sampling and confidence scoring

A note on independence

This course is independently produced exam-preparation content. It is not created, affiliated with, or endorsed by Anthropic. "Claude" and "Anthropic" are trademarks of their respective owner. All exam domain weightings, formats, and structure referenced in this course are provided for study purposes and may change; always confirm current exam details on Anthropic's official certification pages before registering.

CONTENT OVERVIEW / CURRICULUM

Module 1 — Agentic Architecture & Orchestration

1. Agentic Loops and stop reason Explained

2. Multi-Agent Systems and Coordinator Patterns

3. Subagent Context Passing and Session Management

4. Building a Multi-Agent System in Python with the Claude SDK (Hands-On)

5. PreToolUse, PostToolUse Hooks and Task Decomposition

Module 2 — Tool Design & MCP Integration

6. Tool Descriptions and Tool Misrouting Explained

7. Agent Error Handling and tool choice Explained

8. MCP Servers, Configuration, and Client Setup

Module 3 — Claude Code Configuration & Workflows

9. Claude Code Built-in Tools Explained (Grep, Glob, Read, Write, Edit, Bash)

10. CLAUDE Hierarchy and Configuration Rules

11. Custom Slash Commands and Skills

12. Plan Mode vs. Direct Execution

13. Claude Code in CI/CD Pipelines

Module 4 — Prompt Engineering & Structured Output

14. Explicit Criteria Prompting and Reducing False Positives

15. Few-Shot Prompting Explained

16. Structured Output and JSON Schema Design

17. Batch API Usage and Multi-Pass Review Architecture

Module 5 — Context Management & Reliability

18. Why AI Agents Forget: Context Engineering

19. Subagent Error Propagation and Context Management

20. Designing Human Escalation and Review Workflows

Exam Preparation

21. Exam Questions Solved: Official Sample Questions and Common Distractor Traps

22. Exam-Day Playbook and One-Page Cheat Sheet