Claude Certified Architect Practice Exam

Q: How hard is the CCAR-F exam?

Fair if you build with Claude regularly; the closed-book format is what catches people. It assumes working knowledge across five domains, with agentic architecture at 27% and Claude Code at 20%. A full timed mock that lands above the 720 equivalent is the best available predictor.

Q: How long should I study for CCAR-F?

Two to four weeks part-time for engineers already building with Claude: a week of theory against the blueprint, a week of hands-on building with tools and MCP, then timed mocks until the score holds. Longer if the developer platform is new to you.

Q: Is this a real CCAR-F practice exam?

It is a full-length simulation: 60 original questions in 120 minutes, drawn to the published domain weights, with a countdown, a question navigator, flags, and a scored breakdown by domain. The questions are original scenarios, not exam content.

Q: Is the CCAR-F practice exam free?

Yes. No account, no email, no payment. Results are stored in your own browser only.

Q: How is the practice exam scored?

Percent correct is mapped onto the program's 100 to 1,000 scale and compared with the 720 pass mark. Anthropic has not published its scaling, so the score is an estimate of where you stand, not a prediction.

Q: Are these CCAR-F exam dumps?

No. Dumps are leaked exam content, which violates the candidate agreement and gets certifications revoked. These are original questions written from the blueprint. They test the same judgment; they are not the same questions.

Q: What is the difference between the practice exam and the practice questions page?

The practice questions page lets you answer one at a time with instant explanations and filter by domain. The practice exam is timed, hides answers until you finish, and scores you against the pass mark.

Q: How many practice exams should I take before the real CCAR-F?

Until a full timed attempt lands above the pass-mark equivalent with no domain below about 60%. Each attempt draws a fresh domain-weighted set from the bank, so repeats stay useful.

Question 1

A support desk needs incoming tickets sorted into five fixed categories and routed to the matching queue. Which design fits best?

Accepted Answer

A routing workflow with one classification call, then a fixed handoff to the queue for that category

Answer

An orchestrator that spawns five subagents and lets each one claim the tickets it judges to be in scope

Answer

An evaluator-optimizer loop that proposes a category and revises it until a second judge call agrees

Answer

A single autonomous agent with the queue tools attached, deciding per ticket where it should go

Question 2

A task requires editing an unknown number of files across a repository, and the set of files cannot be predicted until the work is under way. Which pattern applies?

Accepted Answer

Orchestrator-workers, with a coordinator that discovers the subtasks during the run and delegates each one

Answer

Prompt chaining with one step per directory, so every part of the repository gets its own fixed pass

Answer

Parallelization with a worker per file type, each scanning its share of the tree for changes to make

Answer

A single call with the full repository loaded into context, asking the model to return every edit in one response

Question 3

A translation pipeline produces drafts that a reviewer can grade against clear criteria, and quality improves when feedback is applied. Which pattern matches?

Accepted Answer

Evaluator-optimizer, where one call drafts, a second grades against the criteria, and the loop repeats

Answer

Parallelization across sentences, with each worker translating a slice and a final merge step

Answer

Routing by language pair, so each pair reaches a specialist prompt tuned with curated examples for that pair

Answer

A single agent with a translation tool that it calls once per document and returns the result

Question 4

An agent can issue refunds. The business wants any refund above $500 approved by a person. Where should that rule live?

Accepted Answer

In a permission check or hook that intercepts the refund tool call and blocks it until a person approves

Answer

In the system prompt, as a rule telling the model to pause and ask before any refund above the limit

Answer

In the refund tool's description, so the model reads the threshold every time it considers the tool

Answer

In a weekly review of the refund log, so any refund over the limit is caught and reversed if it turns out to be wrong

Question 5

A tool that creates invoices sometimes times out, and the agent retries. Customers are receiving duplicate invoices. Best fix?

Accepted Answer

Make the tool idempotent with a client-supplied request id, so a repeated call returns the same invoice

Answer

Raise the tool's timeout so that slow invoice runs finish before the agent loop gives up on the call and retries

Answer

Add a system prompt rule that the model should retry only after checking the invoice list first

Answer

Move invoice creation out of the tool and into a post-run batch job that dedupes by customer

Question 6

During a long multi-agent run one tool starts returning errors on every call, and the coordinator keeps re-issuing the same request. What is missing?

Accepted Answer

Bounded retries with backoff and a circuit breaker that stops calling the tool after repeated failures

Answer

A larger context window for the coordinator so the earlier error results stay visible to it

Answer

A stronger model for the coordinator role so it notices the pattern and changes its approach

Answer

A system prompt instruction telling the coordinator to switch to a different tool as soon as it sees an error result

Question 7

What belongs in a project's CLAUDE.md file?

Accepted Answer

Build and test commands, project conventions, and gotchas Claude needs on every task, kept short

Answer

The full architecture documentation and API references, so Claude has complete knowledge in every session

Answer

Environment variables and service credentials the project needs, so Claude can run commands without asking

Answer

A running log of past sessions and their outcomes, so Claude can pick up where the last one ended

Question 8

A team wants every file Claude Code edits to be run through the formatter, without relying on the model to remember. Which mechanism is right?

Accepted Answer

A PostToolUse hook matched to the edit tools that runs the formatter on the changed file

Answer

A CLAUDE.md instruction asking Claude to run the formatter after each edit it makes

Answer

A custom slash command that formats the tree, which the developer runs at the end of a session

Answer

A formatting subagent that the main session delegates to once the edits are complete

Question 9

How would you run Claude Code non-interactively inside a CI job?

Accepted Answer

Print mode with the -p flag, passing the prompt and a restricted allowed-tools list

Answer

An interactive session on the runner with the permission mode set to accept every action

Answer

A plan-mode session, since planning does not edit files and so needs no approvals

Answer

A subagent launched from a developer's local session, with the CI job waiting on its result

Question 10

A session has been running for hours and answers are getting worse and slower. What is the first thing to try?

Accepted Answer

Compact or clear the context, then re-brief the task with only what the next step needs

Answer

Switch the session to a larger model so it can hold more of the accumulated conversation

Answer

Disable hooks for the rest of the session, since each hook adds latency to every tool call

Answer

Paste the full transcript so far back in as a new user message so nothing is lost

Question 11

You need a response in a strict JSON shape that downstream code will parse. Which approach is most reliable?

Accepted Answer

Define the shape as a tool input schema, or use structured outputs, so the model must conform

Answer

Ask for JSON in the user message and add a note that any text outside the object will break the parser

Answer

Lower the temperature to zero so the formatting is deterministic across runs

Answer

Prefill the assistant turn with an opening brace and stop at the closing brace

Question 12

A prompt includes a 40-page contract and one question about it. Where should the question go?

Accepted Answer

At the end of the user message, after the contract, with the contract clearly delimited by tags

Answer

At the top of the user message, before the contract, so the model knows what to look for while it reads

Answer

In the system prompt, with the contract alone in the user turn, so instruction and content are separated

Answer

Split across several user turns, one section of the contract per turn, with the question at the end

	Real exam	This practice exam
Questions	60	60
Time	120 minutes	120 minutes
Domain weights	Per the published guide	Same weights, applied when the set is drawn
Scoring	100 to 1,000, 720 to pass	Same scale, linear estimate
Format	Proctored, closed-book, no AI	Unproctored; treat it as closed-book to get a real reading
Question bank	Anthropic's, confidential	200 original questions, fresh draw each attempt

The free Claude Certified Architect (CCAR-F) practice exam: 60 questions, 120 minutes.

Take the CCAR-F practice exam

Full practice exam

Quick check

How closely it matches the real CCAR-F

Domain weights used for the draw

How to use it

Take a quick check first

Study the weak domains, then sit the full timed exam

Review every miss, then book when you clear the bar

CCAR-F mock exam, practice test, or practice questions?

What to check in any practice exam

CCAR-F practice exam FAQ

Claude Certified Developer – Foundations

Claude Certified Associate – Foundations

Claude Certified Architect – Professional

Who wrote this

Matei Olaru

Michael Greenberg