Practice exams for the Claude Certified Architect – Foundations exam
Timed mock exams in the exam guide's format, practice that goes after your weak domains, and an honest readiness range. Every question is original, explained and cited. Start free with 10 practice questions a day.
Not affiliated with or endorsed by the exam provider. Exam names identify the exams this material prepares you for.
- questions in each mock exam
- 60
- time limit, as on exam day
- 120 min
- exam domains, all covered
- 5
- exam guide, read 2026-11-09
- v1.0
Try a sample question
One of our own Architect Foundations practice questions, retired so we can show its answer. Written against exam-guide blueprint v1.0, item version 2.
A developer is about to refactor the token refresh logic in a monorepo of roughly 1,800 files using Claude Code. In yesterday's session, a broad search for every caller of the refresh code ran 140 file reads before any edit was made. After that search, per-turn input tokens rose from about 12,000 to about 95,000 and stayed near that level through the refactor. Today the developer needs the same caller list before editing again. Which approach should the developer take for the discovery step?
Choose one.
Show the answer and why
- A. Correct. Broad codebase searches go to the built-in Explore agent. The subagent reads files in its own context window and reports a summary, so the wide search stays out of the main conversation. Asking for short summaries back keeps the results from filling the main context anyway.
- B. Incorrect. The subagent does read files in its own context window. Asking for full file text back brings that bulk into the main conversation anyway. The guidance is to ask for short summaries back so the results don't fill your context.
- C. Incorrect. The reads still happen in the main session, and exploring a large codebase fills your context with file reads. Long output from a wide search rides along for the rest of the session; the documented remedy is to delegate the exploration so only the findings come back.
- D. Incorrect. The file reads still land in the main conversation, where long output from a wide search rides along for the rest of the session. Delegating the exploration instead means only the findings come back.
Sources
Broad codebase searches go to the built-in Explore agent.
Source: cc-best-practicesKeeping long output, like a full test run or a wide search, out of your main conversation, where it would otherwise ride along for the rest of the session
Source: cc-best-practicesThe subagent reads files in its own context window and reports a summary.
Source: cc-common-workflowsExploring a large codebase fills your context with file reads. Delegate the exploration so only the findings come back.
Source: cc-common-workflowsAsk for short summaries back, so twenty results don’t fill your context anyway.
Source: cc-best-practices
Practice inside your AI assistant
With a pass, connect aicertcoach-practice, our Model Context Protocol server, to an MCP client such as Claude Code or Claude Desktop. Our server grades every answer and keeps your progress in the same account as the web app. How to connect it
- Ask for a question. Your assistant fetches the next question from your practice queue, without its answer.
- Answer in the chat. Our server grades it and sends back every option's explanation and its sources.
- Ask how ready you are. Mastery by domain, your readiness range and the reviews due, from the same account as the web app.
Use AI Cert Coach to prepare, never during your exam. The exam provider's policy treats help from AI tools during the exam as misconduct, which can cost you your certification.
Choose your exam
Each pass covers one exam: unlimited practice, timed mock exams and a readiness range, all written from that exam's public guide.
- AvailableExam code CCAR-F
Claude Certified Architect – Foundations
60 questions · 120 minutes
- Coming soon
More exams in preparation
We add an exam once its public exam guide is out and its questions pass review.
Claude Certified Architect – Foundations: the exam at a glance
The Claude Certified Architect – Foundations exam, as its provider describes it:
- Questions
- 60
- Time allowed
- 120 minutes
- Passing score
- 720 on a 100 to 1000 scale
- Exam fee (paid to the exam provider, not to us)
- US$125
- Who can register
- Members of the provider's partner program only
Taken from the provider's public exam guide (version 1.0) on 2026-11-09. Exam details change: verify them with the exam provider before you register.
Free account or Architect – Foundations pass
Free account
Free for as long as you like
- 10 practice questions a day
- One diagnostic every 7 days
- An explanation for every option, with its sources
- Your mastery by exam domain
Architect – Foundations pass
Launch priceUS$19.50 once, for 12 months of access
Free for the first 100 learners while places last, with 30 days of full access; after the free access ends the learner returns to the free tier, and the launch price of US$19.50 applies to a later purchase until 22 January 2027 at 00:00 UTC. One free pass per learner per exam. Checkout shows the exact price before anything is paid.
- No daily limit on practice questions
- Timed mock exams with a score report
- Your readiness as a range with a confidence level
- Spaced review of the questions you missed
- Practice inside your AI assistant, through our MCP server
- Everything in the free account
Refund policy: Ask for a refund within 14 days of purchase and before you start a mock exam, and we refund the amount you paid. A refund ends the pass at once, including any MCP tokens. A free pass has nothing to refund. Free accounts get 10 practice questions a day.
We do not promise a pass. Readiness is shown as a range with a confidence level, never a guarantee. Questions and answers
From diagnostic to exam day
Take the diagnostic
A diagnostic session scores you on each domain of the exam guide, so you know where to start.
Practice what you miss
Practice sessions mix questions from your weakest domains with ones due for spaced review, so the questions you got wrong come back.
Sit a timed mock exam
With a pass, a mock exam has the guide's question count and time limit. Your readiness is shown as a range with a confidence level, never a verdict.
How the questions are made
Every question passes the same six stages, from the public exam guide to your practice queue, and keeps being checked after it ships.
- Stage 1
Official sources
Only the public exam guide and the official documentation, never recalled exam questions.
- Stage 2
AI draft
Each question is drafted to the exam guide and cites its sources.
- Stage 3
Independent review
A separate review checks accuracy and clarity before a question moves on.
- Stage 4
Automated checks
Automated checks confirm every source and catch repeats.
- Stage 5
Human sign-off
A person checks a sample and signs off every release.
- Stage 6
Live and monitored
Every night, learner answers show which questions need another look.
Anything flagged is reviewed again, then fixed or retired.
Original and cited
Written from the official documentation and the public exam guide, never from recalled exam questions. Every question cites its sources.
Versioned to the exam guide
Each question records the exam-guide version it was written against. When the guide changes, affected questions are rewritten or retired.
Recalibrated nightly
Every night we recompute how each question performs from learner answers. Questions that behave oddly are flagged and reviewed.
Start with the diagnostic, free
10 practice questions a day with every option explained. A pass removes the daily limit and adds timed mock exams and your readiness range.