CCA-F Study Strategy¶

Overview¶

This document outlines a 3-phase study approach for the Claude Certified Architect - Foundations (CCA-F) exam. The strategy is designed for someone with 6+ months of production Claude experience who wants to systematically prepare over 4 weeks.

Phase 1 - Foundation Building (Week 1-2)¶

Goal¶

Build a solid understanding of all five exam domains through official documentation and Anthropic Academy courses.

Activities¶

Week 1 - API Fundamentals and Prompt Engineering

1. Complete Anthropic Academy courses:
2. Claude 101
3. AI Fluency Framework and Foundations
4. Building Applications with Claude API

5. Prompt Engineering with Claude

6. Read core documentation:

7. Getting Started - API setup and first request
8. Messages API - Core API reference
9. Prompt Engineering Guide - Official best practices

10. Structured Output - JSON extraction patterns

11. Hands-on practice:

12. Make 20+ API calls with different prompt structures
13. Experiment with system prompts, XML tags, and chain-of-thought
14. Practice extracting structured JSON from unstructured text
15. Build a simple prompt that uses few-shot examples

Week 2 - Tools, MCP, and Claude Code

1. Complete Anthropic Academy courses:
2. Introduction to Model Context Protocol
3. Tool Use with Claude

4. Claude Code Fundamentals

5. Read core documentation:

6. Tool Use Guide - Function calling with Claude
7. MCP Documentation - Protocol fundamentals
8. Claude Code Overview - CLI and IDE integration

9. CLAUDE.md Files - Project configuration

10. Hands-on practice:

11. Define and use 3-5 custom tools via the API
12. Build a simple MCP server with at least 2 tools
13. Set up Claude Code with a CLAUDE.md file
14. Configure custom slash commands and hooks

Key Deliverables for Phase 1¶

• All foundational Anthropic Academy courses completed
• Comfortable making API calls with tools
• Working MCP server
• Claude Code configured for a real project

Phase 2 - Deep Dive and Practice (Week 3)¶

Goal¶

Go deep on the highest-weighted domains (Agentic Architecture at 27%) and build practical experience with production patterns.

Activities¶

Agentic Architecture Deep Dive

1. Complete Anthropic Academy courses:
2. Building Agentic Applications
3. Multi-Agent Systems

4. Production AI Systems

5. Read and study:

6. Agentic Patterns Guide - Official agentic documentation
7. Extended Thinking - Reasoning mode
8. Prompt Caching - Cost optimization

9. Anthropic Cookbook - Agents - Code examples

10. Build an agentic project:

11. Implement a plan-execute-reflect loop
12. Add error handling and recovery
13. Implement a supervisor pattern with multiple sub-agents
14. Measure and optimize cost and latency

Context and Reliability Patterns

1. Study:
2. Long Context Tips - Managing large contexts
3. Rate Limits - API limits and handling
4. Streaming - Real-time responses

5. Error Handling - API error codes and recovery

6. Practice reliability patterns:

7. Implement exponential backoff with jitter
8. Build a validation loop that checks Claude's output
9. Create a fallback chain (Claude 3.5 Sonnet -> Claude 3.5 Haiku)
10. Implement context summarization for long conversations

Key Deliverables for Phase 2¶

• Working agentic application with error handling
• Understanding of all production reliability patterns
• Ability to explain cost/latency tradeoffs for different architectures

Phase 3 - Review and Exam Prep (Week 4)¶

Goal¶

Consolidate knowledge, practice exam-style questions, and fill any remaining gaps.

Activities¶

Practice Exams and Scenarios

1. Work through all scenarios in scenarios.md
2. For each scenario:
3. Read the scenario without looking at the answer
4. Write your solution approach
5. Compare with the provided solution
6. Note any gaps in your understanding
7. Time yourself - aim for under 2 minutes per question

Gap Analysis

1. Review the fact-sheet.md end to end
2. For each concept you are unsure about:
3. Go back to the relevant documentation
4. Build a quick hands-on example
5. Write a brief explanation in your own words
6. Focus extra time on Domain 1 (Agentic Architecture) - it is 27% of the exam

Final Review Checklist

• Can explain when to use agents vs simple prompts
• Can design a multi-agent system with supervisor pattern
• Can configure CLAUDE.md files for a team project
• Can set up hooks and custom slash commands in Claude Code
• Can write effective system prompts with XML tags
• Can extract structured data using tool use
• Can design MCP servers with appropriate tools and resources
• Can explain MCP transport types and when to use each
• Can implement prompt caching for cost optimization
• Can describe reliability patterns for production Claude apps
• Can manage context windows for long-running conversations
• Can handle rate limits and API errors gracefully

Key Deliverables for Phase 3¶

• All practice scenarios completed and reviewed
• No major knowledge gaps remaining
• Confidence in answering scenario-based questions under time pressure

Resources by Priority¶

Must-Have (Primary Study Material)¶

Should-Have (Supplementary)¶

Nice-to-Have (Extra Practice)¶

Exam Day Tactics¶

Before the Exam¶

1. Environment check - Test your ProctorFree setup 24 hours before
2. Quiet space - Online proctoring requires a clean, quiet room
3. ID ready - Have government-issued ID available
4. Rest - Get a full night's sleep. Cramming the night before does not help with scenario-based exams.

During the Exam¶

1. First pass (70 minutes) - Answer all questions you are confident about. Flag uncertain ones.
2. Second pass (40 minutes) - Return to flagged questions. Eliminate wrong answers first.
3. Final review (10 minutes) - Check for any unanswered questions or obvious mistakes.

Question Strategy¶

• Read the full scenario - Important constraints are often in the last sentence
• Identify the domain - Knowing which domain a question targets helps frame your answer
• Look for "best" vs "correct" - Multiple answers may work; choose the most appropriate one
• Production mindset - Anthropic favors reliable, cost-effective, maintainable solutions
• Beware of over-engineering - If a simple prompt works, don't choose the agentic solution
• Watch for MCP vs API tool use confusion - These are different mechanisms for similar goals

Common Pitfalls¶

1. Choosing agents when simple prompts suffice - Agents add complexity. Only use them when multi-step reasoning with tool use is required.
2. Confusing MCP servers and clients - Claude is an MCP client. Your application provides MCP servers.
3. Ignoring cost implications - Extended thinking and large context windows cost more. The exam tests cost awareness.
4. Forgetting prompt caching - For repeated prompts with static prefixes, caching reduces cost by up to 90%.
5. Over-relying on temperature settings - Claude's default temperature (1.0) is usually fine. The exam tests when and why to adjust it.
6. Mixing up tool choice modes - Know the difference between auto, any, and specific tool forcing.
7. Neglecting error handling - Production systems need retry logic, fallbacks, and graceful degradation.
8. Misunderstanding CLAUDE.md precedence - User-level, project-level, and directory-level CLAUDE.md files have specific precedence rules.

Progress Tracking¶

Use the practice-plan.md file for detailed week-by-week tracking with checkboxes.

Milestone Checklist¶

• Phase 1 complete - Foundation courses and docs reviewed
• Phase 2 complete - Agentic project built, reliability patterns understood
• Phase 3 complete - All scenarios practiced, gaps filled
• Exam scheduled
• Exam passed