CCDV-F vs CCAR-F: which Claude exam should you sit, and which first?

CCDV-F vs CCAR-F comes down to one question: do you spend your days building and shipping applications on the Claude API, or designing agent systems, Claude Code workflows and MCP integrations and defending the trade-offs? The first job is CCDV-F: 53 items, eight domains, a third of the paper on applications and integration. The second is CCAR-F: 60 items, five domains, 27 percent on agentic architecture. Same $125 fee, same 120 minutes, same 720 to pass.

CCDV-F vs CCAR-F: the decision table

Nine rows, both codes, nothing inferred. The exams share their logistics almost line for line; the rows that differ should decide it.

Checked against the official sources on 17 September 2026: the CCDV-F certification page, the CCAR-F certification page, their v1.0 exam guides and the partner certification FAQ. One wrinkle before you book: the exam guides give a 24-hour reschedule cutoff, while the Pearson VUE pages and the partner FAQ say 48 hours. Plan for 48.

Who each exam is written for

The Claude developer vs architect certification question is answered by two audience statements. The CCDV-F guide names its audience as AI engineers, technical leads and senior software engineers who turn requirements into working systems. It recommends one to five years of software engineering, six months with Claude or a comparable LLM system, and fluency in Python or TypeScript, REST APIs and the command line. It also says who it is not for: non-technical or casual users, and roles limited to prompt writing.

The CCAR-F guide describes a solution architect who designs and implements production applications with Claude: Agent SDK orchestration, Claude Code configuration for teams, MCP interfaces, structured-output prompting, context management and CI/CD integration. Its items are framed by four scenarios drawn from a published bank of six, and it asks for practical judgment about trade-offs, not only conceptual knowledge.

Read those as job descriptions. The developer owns the code that calls Claude; the architect owns the shape of the system. If you do both, ask which describes more of your week.

Where Claude Code and MCP weigh differently

Five subjects appear on both blueprints, and every one is heavier on CCAR-F. Claude Code is the extreme case: on CCDV-F it is a single skill, Claude Code Operation, at 3.1 percent, roughly two items; on CCAR-F it is Domain 3, Claude Code Configuration and Workflows, at 20 percent, roughly twelve. MCP follows in milder form: Tools and MCPs is 10.6 percent of CCDV-F, Tool Design and MCP Integration is 18 percent of CCAR-F.

The other direction is just as telling. The four CCDV-F domains with no counterpart on the architect blueprint, Applications and Integration at 33.1 percent, Model Selection and Optimization at 16.8, Security and Safety at 8.1 and Eval, Testing, and Debugging at 2.6, add up to 60.6 percent of the developer paper. CCAR-F touches batch cost and credential handling inside other domains, but it has no domain for API mechanics, model tiers, cost or application security.

CCDV-F or CCAR-F first?

If only one job description fits you, sit that exam and stop. The sequencing question arises for the person who both builds and designs, and for them the blueprints answer the cost side.

Going from CCDV-F to CCAR-F, every architect domain has a name you have met, but the weights move sharply: Claude Code grows more than sixfold, context management roughly fourfold, agents and prompting nearly double. Going from CCAR-F to CCDV-F, 61 percent of the paper sits in domains the architect blueprint never names. Neither route is a shortcut, and this post does not choose for you: our guide to which Claude certification fits your role owns that decision.

Which is harder, CCDV-F or CCAR-F?

Nobody can answer honestly, because Anthropic publishes no pass rate, difficulty rating or raw-to-scaled conversion for either exam. Both use a criterion-referenced 720 cut on a 100 to 1,000 scale and allow 120 minutes; shared numbers are logistics, not a difficulty equivalence, which is why how hard the Claude certification is stays personal.

What the blueprints do let you say is how each paper will feel. CCDV-F is the only exam in the programme that publishes a weight for every skill, from Claude Hooks at 1.0 percent to Claude Application Design at 8.6. CCAR-F wraps its 60 items in scenarios you can read in advance, then asks for judgment inside them: which hook, which tool boundary, which escalation rule. An integration engineer and a systems designer will each find a different one of them familiar.

What transfers, and how to practise for either

Prompting and context are named on every blueprint in the programme, so they carry between these papers as they carry everywhere; the cross-track map is in what transfers between the Claude exams. Agents, tools and MCP transfer as vocabulary and judgment, though the same word is a domain on one paper and a skill on the other.

What does not transfer is the material each paper is largest in. The CCDV-F holder arrives at CCAR-F never examined on coordinator and subagent orchestration at 27 percent, or on session state and provenance; the tool side is in tool design and MCP on CCAR-F. The CCAR-F holder arrives at CCDV-F never examined on model tiers, token and cost management, application security or software engineering foundations as domains; their CCDV-F homes are more than half the developer paper.

Both free mock exams are timed and need no account. The free CCDV-F mock is 53 items allocated to the eight-domain blueprint under a 120-minute clock, with raw correct and a per-domain table on submit, never a scaled score or a pass verdict. The free CCAR-F mock is a 20-item set written to the five official domains; it is not built as scenarios and is not a replica of the live paper. Either way, no mock score predicts an official result or tells you to book. Our guide to Claude certification practice exams covers what official material exists, and whether a Claude certification is worth it weighs the fee against what passing changes.

Key takeaways

• Both are $125, 120 minutes and 720 to pass. Shared logistics settle nothing.
• CCDV-F is 53 items over eight domains; CCAR-F is 60 over five. A third of one is applications and integration; a fifth of the other is Claude Code.
• Every shared subject is heavier on CCAR-F. Claude Code goes from 3.1 to 20 percent, tools and MCP from 10.6 to 18.
• 61 percent of CCDV-F has no counterpart domain on CCAR-F. API mechanics, model selection, cost, security and debugging belong to the developer paper.
• Neither is a prerequisite for the other, and difficulty is unpublished. Read the blueprints as job descriptions and pick the one that is yours.