What does the Claude certification cover, and what does it leave out?

What the Claude certification covers depends on the exam: everyday product use for Associate – Foundations; production application building for Developer – Foundations; agentic and tool-centred solution design for Architect – Foundations; and enterprise integration, governance and lifecycle leadership for Architect – Professional. The official blueprints publish five to eight weighted domains and the task statements that define their scope.

A useful study model: one operating cycle

Anthropic publishes domains, weights and task statements; it does not publish a dependency diagram. For study, Cred Farmer maps the Associate scope into one operating loop: choose the product and model, configure the context, prompt, evaluate the output, and place the result in a workflow. Governance constrains every step, while troubleshooting traces a failure back to its cause.

This is an organising device, not a claim about how Anthropic constructs or scores individual items. It helps you revise related decisions together without turning an editorial interpretation into an official exam rule.

What does the Claude certification cover, exam by exam?

The current public pages show four materially different scopes. The counts and highest-weighted domains below were checked on 31 August 2026. Behind the domains sit 121 published objectives, from 25 on CCDV-F to 38 on CCAR-P.

Live source check: CCAO-F, CCDV-F, CCAR-F and CCAR-P.

Why decisions matter more than definition lists

The official FAQ says the exams use multiple-choice and scenario-based multiple-response questions. Their guides then define scope with task statements built around actions such as selecting, evaluating, applying, troubleshooting and designing. Taken together, that supports studying how to make a decision in context instead of memorising feature definitions alone. Eleven of those task statements name a cost, an optimisation or a trade-off outright, six of them on CCAR-P, which is why the option that names what it gave up is so often the credited one.

All four currently use a scaled 720 passing score, even though the forms contain 53, 60 or 63 questions. Anthropic does not publish a raw percentage conversion, so do not turn that shared scaled bar into a claim that the exams have identical difficulty. Prepare the task statements, practise a repeatable evaluation method, and work through sample questions to rehearse the published response formats.

A Cred Farmer method for mixed scenarios

Anthropic does not publish how often an item combines domains or which items are hardest. When one of our practice scenarios appears to mix topics, we use a three-part reading method: identify the decision the stem actually asks for, separate the technical mechanism from the constraint or outcome, then test every response against the stated access boundary, failure mode and goal. This is a Cred Farmer heuristic, not an official scoring rule.

What it deliberately leaves out

The published format contains no live coding lab. All four use multiple-choice and multiple-response questions, so the exam does not directly inspect code you author, a repository you submit or a system you operate during the appointment. You are asked instead to reason about things like a tool-use contract. That is the honest answer when an interviewer would rather see your GitHub. The absence of a lab does not make the technical tracks non-technical: the Developer and Architect blueprints explicitly cover APIs, tools, MCP, agents, Claude Code and system design.

Selected-response testing cannot directly observe typing speed, repository quality or a production delivery record. Nor does the published scope examine collaboration: whether the Claude exams test team skills has a countable answer, one objective in 121, and it is a configuration task on CCAR-P. The format can still ask about Claude Code configuration, API behaviour or cloud deployment choices when the current blueprint lists them. Passing validates performance against that published scope and format; it is not evidence by itself that you have delivered a production system. Keep that distinction clear when you weigh whether a Claude certification is worth it against the fee and the hours, or, if the exams are closed to you today, when weighing preparing without partner network access. If you are still choosing between the four, the guide to which Claude certification to take compares them side by side.

Key takeaways

• Four exams, four role-specific scopes: everyday use, application building, solution design and enterprise-scale architecture.
• The current domain counts are 7, 8, 5 and 7 for CCAO-F, CCDV-F, CCAR-F and CCAR-P.
• The weighted domains and task statements are official. The cycle and role map in this article are Cred Farmer study aids.
• For a mixed practice scenario, separate mechanism from constraint and let the requested decision govern your choice. That is a heuristic, not an exam rule.
• No live build is part of the published format. Technical concepts are still in scope for Developer and Architect candidates.