CCA-F Exam Study Guide: Claude Certified Architect Foundations
▶ Watch on YouTube & subscribe to The Stack Underflow
The CCA-F — Claude Certified Architect Foundations — is Anthropic’s certification for developers building on the Claude API, Claude Code, the Agent SDK, and MCP. If you have been following this series, you already have most of the conceptual groundwork. The exam itself is a structured lens on the same stack you have been studying — and knowing how the domain map lines up with the layers makes your prep a lot more efficient.
This page translates the video’s stack-to-domain mapping into a written reference you can bookmark, highlight, and return to as you work through the episode blocks.
The one-sentence version: Every CCA-F exam domain corresponds to a specific layer or plane of the Claude stack, so weighting your study by domain score percentage is the fastest path to passing.
The Claude Stack Recap
The video references a five-layer mental model called “the Claude stack.” You need this model to understand where each exam domain lives:
┌─────────────────────────────────────────┐
│ L4 Surfaces (Claude Code, etc.) │
├─────────────────────────────────────────┤
│ L3 Orchestration (Agent SDK) │
├─────────────────────────────────────────┤
│ L2 Reach (Tools / MCP) │
├─────────────────────────────────────────┤
│ L1 Protocol (Claude API) │
└─────────────────────────────────────────┘
←── Prompts plane (cuts across) ──→
←── Reliability plane (cuts across)→
Two “planes” — prompts and reliability — are not tied to a single layer. They cut horizontally across every layer, which is exactly why they appear as separate exam domains rather than sub-topics of one layer.
Domain-to-Stack Mapping
| Domain | Topic | Stack Location |
|---|---|---|
| D1 — Agentic and Orchestration | Multi-agent design, orchestration patterns | L1 Protocol + L3 Orchestration |
| D2 — Tool Design and MCP | How Claude reaches external systems | L2 Reach |
| D3 — Claude Code | The canonical developer surface | L4 Surfaces |
| D4 — Prompts and Structured Output | System prompts, output schemas | Prompts plane (all layers) |
| D5 — Context and Reliability | Context windows, retries, evals, safe degradation | Reliability plane (all layers) |
The domain names are not arbitrary branding — they are a direct projection of the stack onto testable competency areas. If you understand the stack, you understand the shape of the exam.
Score Weights: Where to Spend Your Time
This is the most actionable data point in the video. The weights break down roughly as:
- D1 (Agentic and Orchestration): ~27% (referenced as weight D1 27)
- D3 (Claude Code): ~20% (referenced as weight D3 20)
- D4 (Prompts and Structured Output): ~20% (referenced as weight D4 20)
- D2 (Tool Design and MCP): ~18% (referenced as weight D2 18)
- D5 (Context and Reliability): ~15% (referenced as weight D5 15)
The video flags this directly: D1 + D3 + D4 = approximately 67% of the total score. Two-thirds of your points live in those three domains. Weight your study time the same way.
Exam Mechanics
- Questions: 60 multiple-choice questions
- Time: 120 minutes (2 minutes per question — plenty of time if you know the material)
- Scoring: Scaled, 100–1,720; passing score is 720
- Guessing penalty: None — always answer every question
- Retakes: One attempt; verify the current retake policy on Anthropic’s site before you sit
The exam tests architectural judgment, not configuration memorization. Expect production scenarios with four answer options, where two options look plausible and you need to reason about trade-offs rather than recall a flag name.
Your Study Path by Domain
The video maps each domain to specific episode blocks in this series:
| Domain | Episodes to Study |
|---|---|
| D1 — Agentic and Orchestration | Episodes 103, then the 301–306 block |
| D2 — Tool Design and MCP | The 20-series episodes |
| D3 — Claude Code | All of the 5-series |
| D4 — Prompts and Structured Output | All of the 4-series |
| D5 — Context and Reliability | All of the 6-series |
For final practice, work through capstones 702, 703, and 704. These are scenario-based exercises that simulate the architectural judgment questions you will see on the exam.
Common Misconceptions
-
“I can cram config details the night before.” The exam explicitly tests architectural judgment over configuration memorization. Knowing that a flag exists is less useful than knowing why you would choose one pattern over another under production constraints.
-
“Domain 2 (MCP) matters less because it has the lowest weight.” D2 at ~18% is still nearly one-fifth of your score. It also underlies the agent patterns in D1 — skipping it creates gaps in every agent scenario question.
-
“The reliability domain is just about retries.” D5 covers context window management, evaluation design, and safe degradation strategies. It is a systems-thinking domain, not a narrow ops checklist.
-
“Passing the exam means memorizing the series episode list.” The episodes are your study material, not the exam content itself. The exam draws on the concepts those episodes teach, applied to scenarios you have not seen before.
Frequently Asked Questions
Who is the CCA-F for? Developers and architects who build production systems on the Claude API, Claude Code, the Agent SDK, and/or MCP. It is a foundations-level cert — “F” stands for Foundations — so it assumes working familiarity with the API but does not require deep research-track expertise.
Is the exam publicly available? As of June 2026, the program is partner-restricted. Verify current access requirements and exam mechanics directly on the Anthropic site before planning your attempt. Program details evolve, and what was true when this episode was recorded may have changed.
What is the best way to gauge readiness before booking? Use Anthropic’s official practice exam. It is the most accurate signal of whether you are ready — not a third-party quiz bank, and not self-assessment against the domain list. The practice exam also familiarises you with the four-option scenario format before you sit for real.
Should I study the domains in order? Not necessarily by domain number. Study by weight first: D1, then D4, then D3, then D2, then D5. If time is short, that ordering maximises the points you secure before you run out of runway.
Where This Fits in the Series
This video is a course-navigation episode in How Claude Actually Works — it does not introduce new technical concepts but shows you how the concepts you have already learned map to a formal certification framework. Think of it as the syllabus review before the final. If you are new to the series, start from the beginning on the all tutorials page and work through the episode blocks the study path above references. If you are returning specifically to prep for the CCA-F, use the domain table and the episode block map above to slot into whichever series sections cover your weakest domains.
Found this useful? The deep version lives on YouTube — new breakdowns of how AI dev tools actually work, weekly.
Subscribe on YouTube →