Anthropic's certification programme publishes more than one Foundations exam, and two of them are aimed at people who build things. CCDV-F is the Claude Certified Developer – Foundations exam. CCAR-F is the Claude Certified Architect – Foundations exam, the one you will also see written informally as CCA-F — we covered why that exam answers to two different codes separately. The third, CCAO-F, is the Associate exam, written for people who use Claude rather than build with it; we set the Associate exam against the Architect one in its own article.
Both cost $125. Both run 120 minutes. Both are pitched at technical practitioners rather than at general users. So the fee will not tell you which one you are preparing for, and neither will the clock. What separates them is the shape of the exam and the syllabus behind it — and here the two published guides diverge sharply. This article sets them side by side using only what those documents state.
The short answer
The Developer exam is weighted toward getting Claude into an application: the API, the SDKs, the surrounding software engineering, and the design of the app itself. The Architect exam is weighted toward agentic systems and the tooling around them: orchestration, subagents, MCP interfaces, and Claude Code configuration. Both guides describe candidates who write code. They disagree about what that code is mostly doing.
Where the two exams are identical
Set the two at-a-glance tables next to each other — Section 5 of the Developer guide, Section 3 of the Architect guide — and almost every row matches, several of them word for word.
| Detail | CCDV-F (Developer) | CCAR-F (Architect) |
|---|---|---|
| Number of items | 53 | 60 |
| Time limit | 120 minutes | 120 minutes |
| Item format | Multiple-choice and multiple-response; each item states how many responses to select | Multiple-choice and multiple-response; each item states how many responses to select |
| Exam structure | No such row in the guide | 4 scenarios drawn from a bank of 6 |
| Passing score | Scaled score of 720 on a scale of 100–1,000 | Scaled score of 720 on a scale of 100–1,000 |
| Delivery | Proctored: online and/or test centre | Proctored: online and/or test centre |
| Exam fee | $125 USD | $125 USD |
| Validity | 12 months from the date awarded | 12 months from the date awarded |
| Content domains | 8 | 5 |
The item-format row is identical in both guides, down to the punctuation: “Multiple-choice and multiple-response items; each item states how many responses to select”. Neither guide says what proportion of items takes each form, for either exam. That is worth holding onto, because it is an easy thing to assume — we wrote about what multiple-response actually means for how you answer when the Architect guide's wording raised the same question.
The scaled score of 720 on a 100–1,000 range is common to both, and both describe a criterion-referenced assessment measured against a fixed standard rather than against other candidates, with the per-domain percentages on your score report explicitly not used to decide pass or fail. The $125 fee is the same figure in both, which is why it cannot be the deciding factor here.
The policies are not merely similar — the retake sections are word-for-word identical. Both set the same waiting periods after a failed attempt: fourteen days after the first, thirty after the second, ninety after the third, four attempts maximum in a rolling twelve-month period, and the fee payable on each one. Both note the limits apply per exam, so failing one does not stop you registering for the other. Registration for both runs through the Anthropic Partner Academy and Pearson VUE, and both credentials last twelve months and renew through a free, non-proctored assessment.
Where they genuinely differ
The blueprint, and how it is published
This is the substantive difference, and it starts before you read a single domain name. The two guides do not publish their blueprints at the same resolution.
The Architect guide (Section 4) gives five domains in whole percentage points. Its Section 6 lists thirty task statements underneath them, each with “Knowledge of” and “Skills in” bullets, but no weight of its own. You know a domain is worth 20%; you do not know how that 20% is split.
The Developer guide (Section 6) gives eight domains to one decimal place and weights every skill inside them as well — twenty-five skills, each with its own percentage, summing exactly to their domain weights. A candidate can budget study time against it directly.
Here are the eight Developer domains:
| CCDV-F content domain | Weight | Of 53 items |
|---|---|---|
| Agents and Workflows | 14.7% | ~7.8 |
| Applications and Integration | 33.1% | ~17.5 |
| Claude Code | 3.1% | ~1.6 |
| Eval, Testing, and Debugging | 2.6% | ~1.4 |
| Model Selection and Optimization | 16.8% | ~8.9 |
| Prompt and Context Engineering | 11.0% | ~5.8 |
| Security and Safety | 8.1% | ~4.3 |
| Tools and MCPs | 10.6% | ~5.6 |
And the five Architect domains, for comparison:
| CCAR-F content domain | Weight |
|---|---|
| Agentic Architecture & Orchestration | 27% |
| Tool Design & MCP Integration | 18% |
| Claude Code Configuration & Workflows | 20% |
| Prompt Engineering & Structured Output | 20% |
| Context Management & Reliability | 15% |
We break down the five Architect domains and what each one weighs in a separate post.
One domain is a third of the Developer exam
Applications and Integration is 33.1% on its own — roughly 17.5 of the 53 items. Nothing on the Architect blueprint is that concentrated; its heaviest domain, Agentic Architecture & Orchestration, is 27%, or about 16 of its 60 questions.
The Developer guide breaks that domain into six weighted skills: Claude Application Design (8.6%), Software Engineering Foundations (7.4%), Claude API Mechanics (6.8%), Configuration Management (4.1%), Understanding Requirements (3.4%) and Systems Life Cycle (2.8%). Note what is in there — Software Engineering Foundations covers REST APIs, JSON, asynchronous programming, version control and refactoring. That is general engineering practice, not Claude-specific knowledge, and on this exam it alone outweighs two entire content domains.
Two domains round to almost nothing
Claude Code is 3.1% of the Developer exam and Eval, Testing, and Debugging is 2.6% — about 1.6 and 1.4 items. Each is a single undivided skill, and together they come to 5.7%, or roughly three questions. Four individual skills each outweigh the pair: Claude Application Design at 8.6%, Software Engineering Foundations at 7.4%, Claude API Mechanics at 6.8% and Technical Fundamentals at 6.1%.
The contrast with the Architect exam is sharpest on Claude Code: 3.1% of the Developer blueprint against 20% of the Architect one, where it is a domain in its own right covering CLAUDE.md hierarchy, custom slash commands, settings and plan mode. If Claude Code is the reason you are looking at certification at all, that number is the one to read carefully.
The exam's structure — and what the Developer guide does not say
The Architect guide specifies its structure twice: in the Section 3 table as four scenarios drawn from a bank of six, and again in Section 5, which gives all six scenarios a section of their own and names the primary domains each one draws on.
The Developer guide describes no scenario structure at all. Its at-a-glance table has no exam-structure row, and it has no section corresponding to the Architect guide's Section 5. That is a statement about the two documents: one specifies a structure and the other does not. It does not follow that the Developer exam is, or is not, organised into scenarios — the guide simply does not say, and guessing from a silence is how people end up preparing for an exam that does not exist.
The same asymmetry shows up elsewhere. The Architect guide runs to eighteen sections and carries two things the Developer guide has no equivalent of: preparation exercises (Section 8) and an appendix listing technologies and concepts that might appear (Section 17). It publishes twelve sample questions in Section 9. The Developer guide runs to sixteen sections and publishes three samples in Section 8, noting that they illustrate item style rather than coming from the operational pool.
If you cross-reference the two documents, be careful with section numbers: they do not line up, and the offset is not constant. Exam details are Section 5 in the Developer guide against Section 3 in the Architect one, and the blueprint Section 6 against Section 4 — but registration is Section 10 against Section 11, and the confidentiality agreement Section 13 against Section 14.
Who each exam is written for
The Developer guide's Section 3 is specific. It names AI and machine learning engineers, technical leads and senior software engineers, and expects one to five years of software engineering experience, at least six months hands-on with Claude, proficiency in Python and/or TypeScript, and fluency with REST APIs and command-line tools. It then says plainly who the exam is not for: non-technical or casual users, anyone without hands-on software development experience, and roles limited to prompt writing or other isolated tasks.
The Architect guide's Section 2 describes a solution architect who designs and implements production applications with Claude, with hands-on experience across the Agent SDK, Claude Code, MCP interfaces and context management, and typically six or more months of practical work against those technologies. Both profiles assume a working engineer; the Developer one is framed around shipping an application, the Architect one around designing the system's agentic structure. Our longer piece on which roles the Architect certification actually fits works through that side in more detail.
Neither guide requires anything. The Developer guide's Section 4 states there are no mandatory prerequisites or courses, and that the credential is awarded on exam performance alone.
How to choose
- What are you mostly building? If it is an application that calls Claude — API integration, streaming, error handling, model and cost decisions — the Developer blueprint puts half its weight there. If it is an agentic system with subagents, orchestration and MCP tooling, the Architect blueprint is built around that.
- How much does Claude Code matter to you? 20% of one exam, 3.1% of the other. This is the single largest divergence between the two blueprints.
- Do you want the syllabus at skill resolution? The Developer guide gives you twenty-five weighted skills; the Architect guide gives thirty task statements with no weights attached. If you plan study time against percentages, that is practical rather than cosmetic.
- Are you comparing on price? You cannot. Both are $125, with the same retake costs and waiting periods. The decision has to be made on content.
Nothing in either guide ranks one credential above the other, and neither is described as a prerequisite for the other. They are different scopes of work with different blueprints, and nothing stops you holding both.
What we cover, and what we do not
Being straightforward about this: our practice material is written for the Architect exam, CCAR-F. We do not offer CCDV-F practice questions. If you came here looking for Claude Certified Developer practice tests, we do not have them, and our Architect material is not a substitute — the blueprints overlap in subject matter but not in weighting, and an exam where Claude Code is 3.1% is not well served by a bank built for one where it is 20%.
If the Architect exam is the one you are preparing for, our guide to the exam's format, domains and scoring is the place to start, and the free readiness diagnostic gives you a per-domain score before you commit to anything. The full CCAR-F practice question bank covers all five Architect domains.
CCA Practice Platforms is an independent exam-preparation resource. We are not affiliated with, endorsed by, or otherwise connected to Anthropic, and we do not administer or register anyone for the official exams. Every exam fact in this article comes from the published Claude Certified Developer – Foundations and Claude Certified Architect – Foundations exam guides, both version 1.0, effective July 2026. Both guides state that they are subject to change without notice, so verify current details on the Anthropic Partner Academy before you register.