Claude Certified Architect Foundations (CCA-F) – Study Plan

Overview

The CCA-F validates your ability to design and deploy production-grade Claude integrations. It tests architectural judgment in real-world scenarios, not definition recall. Questions are anchored in system design problems where you must choose the right approach, justify trade-offs, and reason about reliability, cost, and safety.

If you already work with the Claude API and Claude Code, you have the hands-on foundation this exam assumes. This plan builds on it by shifting your focus from "how do I use this feature?" to "when and why is this the right architectural choice?"

Agentic architecture is the single largest domain and the one most candidates underestimate. It is also the conceptual foundation for how questions in other domains are framed. This plan dedicates two weeks to it.

Recommended experience: Approximately 6 months of hands-on experience with the Claude API and Claude Code. Understanding of LLM APIs, prompting fundamentals, and message roles.

Frequently Asked Questions

Do I need prior LLM experience?

Basic understanding of LLM APIs is helpful but not required. The certification covers Claude-specific concepts. Approximately 6 months of hands-on experience with the Claude API and Claude Code is a realistic baseline, though motivated candidates with less experience have passed with focused preparation.

Why does this plan dedicate two weeks to agentic architecture?

Because it is 27% of the exam and the conceptual foundation for every other domain. Tool design, context management, and Claude Code configuration all assume you understand agentic patterns. Candidates who skim this domain consistently underperform across the entire exam.

What if i have only 4 weeks to study instead of 6 weeks?

• Week 1: Merge current Weeks 1–2 (all of Agentic Architecture). This is the biggest domain but do not shortcut it, just move faster.
• Week 2: Merge current Weeks 3–4 (Tool Design/MCP + Claude Code Configuration). These overlap conceptually (MCP servers show up inside Claude Code), so studying them back-to-back reinforces rather than fragments.
• Week 3: Current Week 5 unchanged (Prompt Engineering + Context Management).
• Week 4: Current Week 6 unchanged (Legend mode + booking). Do not compress this, it is the gate that tells you whether you are actually ready.

How many practice questions should I do?

Aim for at least 600 unique questions across all domains, ensuring you prioritize quality over quantity. ExamOS provides over 2,000 high quality practice questions covering all domains and designed to explain the rationale behind both correct and incorrect answers.

When should I book the exam?

Once you are consistently hitting 80% or above on Legend mode across two or more sessions, and you feel confident explaining architecture decisions for production Claude systems, you are ready. Trust your preparation.

How long is the exam?

120 minutes for 60 questions. That averages to 2 minutes per question, but scenario-heavy questions with long context may take 3-4 minutes. Budget your time accordingly.

What are the core technologies tested?

Claude Code, Claude Agent SDK, Claude API, and Model Context Protocol (MCP).

How is this different from other AI certifications?

CCA-F focuses on decision-making in production contexts, not definition recall. Every question is anchored in a scenario where you must choose the right architectural approach. Knowing what a feature does is table stakes. Knowing when to use it and when not to is what the exam measures.

What is the most important domain to study?

Agentic architecture and orchestration (27%) is the single largest domain and the conceptual foundation for how questions in other domains are framed. Tool design, context management, and Claude Code configuration all build on agentic patterns. Get this domain right first.

Can I use reference materials during the exam?

No. The CCA-F exam is closed-book. You cannot access external documentation, websites, or AI assistants during the test. All questions are designed to be answerable from your knowledge of Claude architecture, the Claude API, Claude Code, and MCP. Study and practice thoroughly before exam day.

What are the most common ways people fail?

• Confusing agents with workflows. Knowing when agentic architecture adds value versus when a deterministic pattern is simpler and more reliable is tested constantly.
• Ignoring token cost trade-offs in multi-agent systems. Parallel subagent dispatch is powerful but expensive. The exam tests whether you understand when the cost is justified.
• Poor tool descriptions. Vague descriptions cause the model to misroute or fail to select the right tool. This is one of the most common failure scenarios.
• Not understanding context rot. Assuming more context is always better is a trap. Long-running sessions degrade accuracy, and the exam tests whether you know how to fix it.
• Studying features in isolation. The exam tests combined scenarios where multiple domains interact: an agent with MCP tools, structured output, and context management all in one question.

What are the three biggest preparation mistakes?

1. Studying documentation without building. CCA-F is a practical, scenario-based exam. If you have not built an agent loop, stood up an MCP server, or configured CLAUDE.md precedence, you will struggle with scenario questions.
2. Skipping agentic architecture depth. It is 27% of the exam and the foundation for every other domain. Candidates who skim this domain consistently underperform.
3. Not practicing under time pressure. 60 questions in 120 minutes means you cannot afford to deliberate on every question. Practice timed quizzes to build decision-making speed.

Quick-Reference: What Is on the Exam (Exam Guide v0.2, June 30, 2026)

This is a condensed summary. For the definitive list, visit Anthropic's certification page.