FREE BONUS
Top 25Claude CertifiedArchitectFoundations Exam Questions
Practise for the Claude Certified Architect, Foundations exam. Every question comes with the correct answer, why it is right, and why each of the other options is wrong.
- 25 exam-style questions across all 5 exam domains
- A full explanation for every option, right and wrong
- Medium and Hard tags, plus a domain map to find your weak spots
Get the free question bank
Fill in your details and we will send the PDF straight to your inbox.
- 25Exam-style questions
- 5Exam domains covered
- 10Hard scenario questions
- 75Wrong options explained
All five exam domains, covered
Questions are spread across the five domains of the Claude Certified Architect, Foundations exam.
Agentic architecture & orchestration
Agentic loops, hub-and-spoke agents, hooks and escalation.
Tool design & MCP
MCP, tool descriptions, secrets and structured errors.
Claude Code config & workflows
CLAUDE.md, headless mode, Plan Mode and skills.
Prompting & structured output
Explicit criteria, JSON schemas, few-shot and retries.
Context management & reliability
Long context, batches, checkpoints and provenance.
Claude Certified Architect Questions
How do you make Claude Code follow the team's coding standards in every session in a project?
Which flag runs Claude Code non-interactively in a CI/CD pipeline?
Why should code generation and code review run in separate sessions?
explanations + 24 more questions
More than an answer key
The wrong options are the traps the exam sets. This guide explains every one of them.
- Exam domain on every questionA domain map shows which questions test each of the five domains.
- Difficulty rating15 Medium and 10 Hard questions, including scenarios where two options look right.
- Why it is right, why the rest are wrongLearn the reasoning, so you can handle new questions on exam day.
Try a Hard question now
Here is question 10 from the guide. Pick your answer first, then reveal the explanation.
No refund can exceed $500 without manager approval. Option 1: add "never approve refunds over $500" to the system prompt. Option 2: add a hook that blocks any refund tool call over $500 before it runs. Which is correct?
Reveal the answer
Why B is correct
Hooks enforce rules deterministically; prompts guide behaviour probabilistically. A PreToolUse hook runs before the refund tool executes, inspects the amount and blocks the call when it exceeds $500, every time. It has to be PreToolUse: a PostToolUse hook runs after the tool, when the refund has already been issued.
Why the others are wrong
- A: System prompt text is not binding. It is input the model follows most of the time, not always.
- C: They are not equally reliable. For a financial limit, the difference between code and a prompt matters.
- D: The system prompt does not override tools. Tools run independently; the prompt is context the model reasons about.
Three steps to get exam ready
Pick your answer first
Read the question and choose an option before you look at the highlighted answer. That is how the exam works.
Read why, not just what
Read why the answer is correct and why each other option is wrong. The wrong options are the traps the exam sets.
Mark your weak domains
Note every question you got wrong and its domain. Those domains are where to study next.
Learners who prepared with K21 Academy
Hear it from K21 Academy learners who built hands-on AI skills, passed certifications and got hired.
“I'm thrilled to share that I've recently landed a role as a Generative AI Engineer.”The hands-on projects let me confidently describe an end-to-end AI solution in the interview. The interviewer was genuinely impressed.
“I have finally secured a contract-to-hire role as a Generative AI Engineer.”I was deploying a RAG pipeline to a web app during the actual hiring process. The projects I built at K21 were exactly what the company needed.
“I've successfully landed a job as an AI Engineer.”Worked hands-on with Azure AI services, Azure AI Foundry, and OpenAI. Later received a second offer as a Generative AI Engineer. K21 gave me the skills and confidence to choose.
“I got cracked that role.”The Azure AI/ML trainee landed the job after completing the program and building out her portfolio.
“I've secured a position as an OpenShift Architect.”Credits the structured training and support system for making the transition into a cloud architecture role.
“Enrolling in K21 Academy's AI/ML program has been one of the best professional decisions I've made this year.”As a VP of Software Engineering transitioning into AI/ML, I completed four courses and passed three Microsoft certification exams: AI-900, AZ-900, and AB-731. The hands-on projects on Azure gave me practical experience that actually shows up in interviews.
Before you download
Is the guide really free?
Yes. Fill in the form and the PDF is sent to your inbox at no cost.
Are these the official exam questions?
No. They are practice questions written by K21 Academy to help you prepare, in the style of the exam. K21 Academy is not affiliated with Anthropic.
Does it include answers?
Yes. Every question shows the correct answer, why it is right, and why each of the other options is wrong.
Which topics does it cover?
All five exam domains: agentic architecture and orchestration, tool design and MCP, Claude Code configuration and workflows, prompting and structured output, and context management and reliability.
Who is this guide for?
Developers, architects, testers, data engineers, project managers, UI/UX designers and business analysts preparing for the Claude Certified Architect, Foundations exam or using Claude at work.
Want to pass the exam and use Claude at work?
These questions are the first step. K21 Academy's hands-on Claude AI for Agentic Development bootcamp takes you from Claude basics to building agentic workflows with Claude Code.
Download the free questions12 hands-on modules
Agentic AI workflows with Claude Code, from setup to real-world labs.
Prompting for reliable agents
Prompt engineering, scalable agentic architectures and engineering standards.
Enterprise ready
Testing, DevOps integration, governance, security and responsible AI.
Bonus courses included
AWS Cloud Practitioner, Azure AZ-900, Python and Azure AI Fundamentals AI-901.