Claude Certified Architect - Foundations

A Claude Certified Architect is an engineer who has passed Anthropic's CCA-F exam, a proctored 120-minute test with 60 scored questions on designing and shipping production Claude applications. The certification validates skills in agent architecture, multi-agent orchestration, MCP integration, prompt engineering with structured output, and context management. Anthropic runs the program through its Partner Academy with delivery by Pearson VUE, and the credential stays valid for 12 months.

For engineers who build on Claude, it is currently the clearest way to show verified, up-to-date skills. The exam tests the production topics teams hire for today: agentic systems, MCP integration, Claude Code workflows, and reliability patterns. Because the credential renews every 12 months, holding it signals that your knowledge is current. Preparing for CCA-F also builds skills you will use on the job, so the study time pays off even before you add the badge to your profile.

CCA-F has 60 scored questions with a 120-minute time limit, which works out to about two minutes per question. Questions are scenario-based and test architectural judgment across five domains: Agentic Architecture & Orchestration (27%), Claude Code Configuration & Workflows (20%), Prompt Engineering & Structured Output (20%), Tool Design & MCP Integration (18%), and Context Management & Reliability (15%).

The passing score is 720 on a scaled 0-1000 scale. Scaled scoring means the raw number of correct answers is converted to account for small differences between exam forms, so there is no fixed count of questions you must get right. On practice tests, aim for a consistent 75-80% before booking.

The exam fee is $125 USD as of the July 2026 program update (it was $99 during the initial Skilljar-delivered period). Registration goes through the Anthropic Partner Academy, which requires free membership in the Claude Partner Network; the exam itself is delivered by Pearson VUE, either online proctored or at a test center.

The credential is valid for 1 year. Anthropic offers a free, non-proctored renewal assessment on the Partner Academy before the 12-month expiration; if the credential lapses, a full proctored retake is required.

No. Anthropic lists no prerequisites for any of its four certifications, and CCA-F is not required for the Professional-tier CCAR-P. CCA-F is the natural first step if you are still building daily implementation skills with the Claude API, Claude Code, the Agent SDK, and MCP; CCAR-P assumes that vocabulary and tests architect-level judgment on top of it.

CCA-F is harder than entry-level certifications because it tests decisions rather than definitions: which tool_choice setting, how to scope CLAUDE.md files, when to use the Batch API, how to isolate subagent context. On Preporato practice tests, the first-attempt median score is in the low 80s and Prompt Engineering & Structured Output is the weakest domain, so plan hands-on time with the tools you will be asked to configure.

Anthropic recommends about six months of hands-on production experience with the Claude API, Claude Code, the Agent SDK, and MCP. Candidates who have built at least one multi-agent system, one MCP server, and one structured-output pipeline are well placed; the exam rewards having configured these things yourself rather than having read about them.

Six full-length practice tests (390+ scenario questions) with an explanation for every option, a 500-card spaced-repetition flashcard deck, hands-on build-and-submit projects graded against a rubric, and a free 20-question sampler you can try before buying. Practice tests are $19.99 one-time with lifetime access, or included in Preporato Pro alongside every other certification and lab.

Yes. CCA-F has no formal prerequisites, so anyone can register and take the exam. Anthropic recommends about six months of hands-on production experience with the Claude API, Claude Code, the Agent SDK, and MCP before you sit it. Candidates coming from other AI stacks can close that gap with focused practice on agents, MCP servers, and structured output, which is exactly what the exam weights most heavily.