CCAR-F vs CCA-F: Which Is the Real Claude Certified Architect Exam Code?

Same certification, one official code

The credential's full name is Claude Certified Architect – Foundations. The community compressed that to CCA-F early on, and the abbreviation spread through course titles, GitHub repos, and forum posts. Anthropic's own surfaces never used it.

The official code — the one in the exam guide, on registration, and on your score report — is CCAR-F. If you hold material that says CCA-F, it is not automatically outdated; it just uses the informal name. What you should check is whether it covers the current item formats, because that is where older community material quietly falls behind.

Practically: search for both terms while you prepare. Most of the good community discussion still lives under CCA-F, while official logistics live under CCAR-F. Study platforms, including ours, anchor both names precisely so candidates coming from either direction end up in the same place.

What the official guide specifies: the blueprint

Anthropic's Exam Guide v1.0 (effective July 2026) defines the content blueprint. Five domains carry the weights: Agentic Architecture & Orchestration at 27 percent, Tool Design & MCP Integration at 18 percent, Claude Code Configuration & Workflows at 20 percent, Prompt Engineering & Structured Output at 20 percent, and Context Management & Reliability at 15 percent. Detailed task statements sit under each domain, and six named scenarios frame the questions — with each exam form presenting 4 scenarios drawn at random from that bank of 6.

The mechanics: sixty items, 120 minutes, scaled scoring from 100 to 1,000 with 720 required to pass, and $125 USD per attempt (partner-tier discounts, where applicable, are reflected at checkout). Scoring is criterion-referenced — you pass by meeting a fixed standard, not by beating other candidates, and Anthropic does not publish a raw-percentage conversion. Your score report shows pass/fail, your scaled score, and percent-correct per domain.

Every hour you spend on agentic architecture patterns, tool schema design, Claude Code configuration, prompt structure, or context management maps directly onto this blueprint, whichever name your study material uses.

The format detail older material misses: multiple-response items

The guide states that items are multiple choice AND multiple response, and that each item tells you how many responses to select — you will see phrasing like "Select TWO" in the question stem.

This matters more than it sounds. Multiple-response items punish partial knowledge much harder than single-answer items. With one correct answer, recognising the best option is enough. With select-two, you must independently validate each option, because one confident pick plus one plausible-sounding distractor is a wrong answer. The classic failure mode is anchoring on the obviously correct choice and then grabbing the most familiar-looking second option instead of eliminating rigorously.

Anthropic has not published how many of the sixty items are multiple-response. Treat any source claiming an exact count as guessing. The defensible preparation strategy is format-agnostic rigour: for every option in every practice question, decide true-or-false on its own merits before looking at the answer combination. That habit transfers to whatever mix the live exam deals you.

Our practice bank includes 120 dedicated multiple-response questions across all five domains, each stating how many responses to select, exactly as the guide describes — because practising the format cold on exam day is the avoidable mistake.

Logistics: delivery, policies, and retakes

The exam is proctored and delivered through Pearson VUE — either at a test centre or from home via online proctoring. Registration runs through Anthropic's Partner Academy: review the exam details there, register, then schedule your session through Pearson VUE.

Exam-day policies are standard proctored-certification fare, and worth knowing in advance: you must present valid government-issued photo ID exactly matching your registration name (corrections go to certifications-support@anthropic.com before scheduling), you must accept a non-disclosure agreement before the exam starts, and accommodations must be requested and approved through Pearson VUE before you book. You can reschedule or cancel up to 24 hours before your appointment; changes inside 24 hours forfeit the fee. Decisions can be appealed within 14 days.

Retake policy follows a graduated wait: 14 days after a first fail, 30 after a second, 90 after a third, with a maximum of four attempts in any rolling twelve months, at $125 per attempt. The credential is valid for twelve months; renewing on time is a free, non-proctored assessment on the Partner Academy, while a lapsed credential requires a full retake at the full fee.

How to rehearse the real format

The exam guide itself is the first preparation layer: its task statements double as a self-audit checklist — if you cannot sketch a competent answer to a task statement, that is a study gap. It also includes a set of sample questions with explanations, drawn from the same style as the live exam, which show you exactly how scenario-grounded the items are.

The second layer is hands-on work: building even a small tool-using agent teaches context management and tool design more durably than any flashcard.

The third layer is timed simulation under exam conditions. Whatever you use, insist on two things: scenario-based questions rather than definition recall, because the real exam places you inside architectural decisions; and multiple-response coverage that states how many answers to select, because that is the one skill older single-answer material cannot give you. Our simulator replicates the 60-question, 120-minute, 720-to-pass structure with both item formats and the 4-of-6 scenario draw — and the first fifteen questions are free, so you can judge the fit before paying anything.

What this means for your study plan

If you are four weeks or less from your exam date, change almost nothing. Your domain-by-domain plan stands. Add one adjustment: convert a slice of your practice time to multiple-response drills, and score yourself strictly — a select-two item with one right pick is zero, not half.

If you are starting fresh, build the plan against the current guide from day one. Weight your hours roughly like the exam does: about a quarter on agentic architecture and orchestration, a fifth each on Claude Code and prompt engineering, and the remainder split between tool design with MCP and context management. Work scenario-first — for each task statement, ask what could go wrong in production and what the disciplined architectural answer is — because that is the shape of the questions.

Either way, confirm your registration path on the Partner Academy early, book the Pearson VUE slot before your motivation dips, and rehearse at least one full 120-minute simulation in the final week. Candidates who walk in having already felt the time pressure, the format mix, and the per-domain scoring consistently report calmer exams than candidates who studied the same hours without simulation. The two-name confusion cost the community some clarity; it should not cost you any points.