Claude Certified Architect – Foundations: Frequently Asked Questions

CCA-F validates whether you can design and reason about production agentic applications built with Claude at a Foundations level of depth. The exam is organized into five domains: Agentic Architecture & Orchestration (27% — multi-agent orchestration, subagent delegation, tool integration, and lifecycle hooks using the Claude Agent SDK), Claude Code Configuration & Workflows (20% — CLAUDE.md, Agent Skills, MCP server integrations, and plan mode for team use), Prompt Engineering & Structured Output (20% — engineering prompts that reliably produce structured, schema-conformant output), Tool Design & MCP Integration (18% — designing MCP tool and resource interfaces, not just consuming existing ones), and Context Management & Reliability (15% — managing context windows across long documents and multi-agent handoffs, plus escalation and reliability decisions). It is delivered through 60 scored, multiple-choice/multiple-response items built around four narrative scenarios drawn from a pool of six, so it tests applied judgment on realistic architecture problems rather than isolated recall of facts.

The exam consists of 60 scored items, delivered as multiple-choice and multiple-response questions anchored to four scenarios selected from a pool of six possible scenarios. You have 120 minutes to complete it. Scoring is reported on a scaled range of 100 to 1,000, with a passing score of 720. Your result report shows the pass/fail outcome, your scaled score, and a percent-correct breakdown by domain — the domain breakdown is genuinely useful for guiding further study even after you pass, since it shows where your margin was thin versus strong. On simple time math, 120 minutes across 60 items works out to an average of 2 minutes per item, but the scenario-based structure means that budget isn't evenly spread — you'll likely spend more time reading and orienting yourself in each of the four scenarios up front, then move faster through the several questions anchored to each one once the scenario's context is clear in your head. Pacing practice under a timer, not just content review, is worth building into your prep specifically because of this structure.

It's calibrated to be genuinely challenging for someone without real hands-on experience, and manageable for someone who matches the intended candidate profile. The exam is scenario-based rather than fact-recall-based, which means memorizing definitions alone won't carry you — you need to have actually made architecture decisions like the ones described in each scenario (choosing between single-agent and multi-agent delegation, deciding MCP tool granularity, debugging a context-window problem) to reliably pick the correct answer among several plausible-sounding options. Anthropic's own guidance points to roughly 6+ months of hands-on experience with Claude APIs, the Agent SDK, Claude Code, and MCP as the level the exam is designed for. If that describes your background, the exam is difficult in the way a well-designed professional exam should be — testing judgment under time pressure — rather than difficult because of trick questions or obscure trivia. See [eligibility](/claude-certified-architect/eligibility) for a detailed self-assessment.

There's no single honest number, because it depends entirely on your starting point, and Anthropic doesn't publish a specific study-hours recommendation. A candidate who's already built multi-agent systems with the Agent SDK, configured Claude Code for a real team, shipped an MCP server, and debugged context-window issues in production may only need a handful of hours to review the domain blueprint and confirm coverage before scheduling. A candidate earlier in that journey should expect a materially larger investment — and, more importantly, should treat the gap as something to close with real hands-on projects (see the concrete project suggestions on the [eligibility page](/claude-certified-architect/eligibility)) rather than pure reading, because the scenario-based format rewards applied experience over memorized facts.

$125 USD per attempt. That covers one sitting of the 120-minute, 60-item exam. There's no discount for retakes — if you don't pass, the next attempt costs the same $125 again, on top of the mandatory waiting period before you're eligible to retake. The $125 is the only guaranteed cost; it doesn't include any paid third-party prep materials, and it doesn't include the value of your own prep time, which for most candidates is the larger real cost of pursuing this credential. Displayed checkout pricing reflects standard currency conversion and any partner-tier discount that applies automatically — there's no separate manual discount code to enter. Full cost and reimbursement guidance, including how to build an employer reimbursement request and what to do if your company needs a formal invoice rather than a receipt, is on the [cost and registration page](/claude-certified-architect/cost-registration).

Registration happens exclusively through the Anthropic Partner Academy, with scheduling and delivery handled by Pearson VUE (online-proctored or test-center options). The process runs: review the exam page on the Partner Academy, download the official exam guide and read the certification terms, register and complete checkout, create your Pearson VUE account from the confirmation instructions, then schedule your session. The "access request" candidates search for typically refers to this account-creation and sign-in step on the Anthropic Partner Academy — you generally need a Partner Academy account (which may involve a brief access or account-approval step depending on your partner status) before you can complete registration and checkout. There is no separate official channel outside the Anthropic Partner Academy and Pearson VUE — treat any third-party site offering registration, vouchers, or guaranteed passes outside this flow as unofficial and untrustworthy. Full step-by-step detail lives on the [shared policy reference](/claude-certifications/policies).

There are no mandatory prerequisites. Anyone can register and sit the exam regardless of prior certifications, degrees, or job title. What Anthropic publishes instead is a minimally qualified candidate (MQC) profile — a description of the experience level the exam is calibrated for, roughly 6+ months of hands-on work with Claude APIs, the Agent SDK, Claude Code, and MCP, building agentic applications with multi-agent orchestration, subagent delegation, and lifecycle hooks. That's guidance about readiness, not a registration gate. The full eligibility breakdown, including a concrete self-assessment and what to build first if you're not ready yet, is on the [eligibility page](/claude-certified-architect/eligibility).

You can retake the exam, but only after a waiting period that increases with each failed attempt: 14 days after a first failed attempt, 30 days after a second, and 90 days after a third. There's a maximum of 4 total attempts within any rolling 12-month period. Each attempt costs the full $125 fee again — there's no reduced retake price. Practically, that escalating structure means your second attempt is the one where preparation quality matters most: a 14-day gap is short enough that you can realistically close a narrow gap (say, one weak domain flagged by your score report) and try again quickly, but a second failure pushes you into a 30-day wait, and a third into 90 days — at which point the compounding cost in both money and elapsed time is significant enough that treating each attempt as disposable stops making sense. These waiting periods and attempt limits are identical across all four Anthropic Claude certifications and are documented in full on the [shared certification policies page](/claude-certifications/policies).

Yes — it's valid for 12 months from the date it's awarded, not from your registration or exam date. Before it expires, you can complete a free, non-proctored renewal assessment on the Anthropic Partner Academy covering what's changed since you certified; there's no fee for on-time renewal. If the credential lapses before you renew, you must retake the full exam at the full $125 fee to regain certified status. Anthropic may also require a full retake instead of the lighter renewal assessment if exam content has changed significantly since you certified. A full worked timeline example, a lapse-consequences comparison table, and guidance on reading your score report are on the [validity page](/claude-certified-architect/validity).

Your result report includes the pass/fail outcome, your scaled score out of the 100-1,000 range against the 720 passing bar, and a percent-correct breakdown by each of the five exam domains. That domain breakdown is worth using deliberately: if you passed, it shows you where your margin was strongest and weakest, which is a better guide for continued learning than the pass/fail headline alone. If you didn't pass, it's the most efficient way to target your mandatory waiting-period study time — concentrate on the specific domain or domains where the breakdown shows the biggest gap, rather than re-studying everything uniformly.

The single authoritative resource is Anthropic's own official exam guide for CCAR-F, available through the Anthropic Partner Academy — it documents the exact domain weights, task statements, and knowledge/skills expectations the exam is built from, and should be your starting reference before anything else. Beyond that, hands-on building is more valuable than passive reading for this particular exam, given its scenario-based format: real practice building Agent SDK orchestration, Claude Code configuration, and MCP integrations will serve you better than a study guide alone. Third-party prep resources (practice questions, mock exams, walkthroughs) can be a reasonable supplement for checking your understanding under exam-like conditions, but treat the official exam guide as ground truth and be skeptical of any third-party resource that claims more authority than that.

No legitimate source of leaked or "real" CCA-F exam questions exists, and you should treat any site claiming to sell actual exam content with real skepticism. Using genuine leaked exam material, if it existed, would also almost certainly violate the certification terms and conditions you agree to when registering, putting your credential at risk even if you passed using it. What you'll typically find under "exam dumps" searches is either outdated, fabricated, or low-quality unofficial practice content mislabeled as real exam questions — which is a poor substitute for genuine preparation, since the scenario-based format specifically rewards applied understanding over memorized answer patterns. The legitimate alternative is practicing with well-constructed mock questions and scenario walkthroughs built from the same domain blueprint Anthropic publishes — that's the approach this site's own practice questions and mock exams are built around, and it's a far more durable way to prepare than searching for dumps.

No — the same concerns apply, and forum-sourced material carries additional risk because it's unmoderated and unverifiable. Reddit threads discussing CCA-F prep can be a reasonable place to gauge general sentiment about difficulty or pacing, but any thread claiming to share actual exam questions or answers should be treated as unreliable at best and a policy violation risk at worst if you use it as your primary prep method. There's also a practical problem independent of the ethics: forum-sourced "dumps" are frequently wrong, outdated relative to the current exam version, or simply fabricated, so relying on them risks learning incorrect information that actively hurts your real exam performance.

Sites like ExamTopics aggregate community-submitted questions across many certifications, and the accuracy, currency, and legitimacy of any specific certification's listings there varies and isn't something this page can verify or vouch for. Given the same concerns that apply to exam dumps generally — unverified sourcing, potential policy violations, and the risk of studying material that's simply wrong — the safer path is to prepare from Anthropic's official exam guide and from practice content that's explicitly built as original mock material against the published domain blueprint, rather than community-aggregated question banks of uncertain origin. A useful test to apply to any third-party question bank, whether ExamTopics or elsewhere: does it cite the actual domain names and weights from the current official exam guide (Agentic Architecture & Orchestration at 27%, and so on), or does it use vague, generic framing that could apply to almost any AI certification? Generic framing is a strong signal the content wasn't built specifically against CCA-F's actual blueprint, regardless of what it claims.

Yes. "CCA-F" is the colloquial shorthand people commonly search for and use in conversation; the certification's official full name is Claude Certified Architect – Foundations, and its official exam code, as published in Anthropic's exam guide, is CCAR-F. All three refer to the exact same credential — there's no separate or different certification hiding behind the different names. This page and the rest of this site use all three terms interchangeably for that reason.

Foundations and Professional are two distinct tiers within the Architect certification track, with different exam guides, domain blueprints, and fee structures. Foundations (this exam, CCAR-F/CCA-F) tests five domains — agentic orchestration, Claude Code configuration, prompt engineering, MCP tool design, and context management — at a working-competency level for solution architects with 6+ months of hands-on Claude/Agent SDK experience. Professional (CCAR-P) tests a broader seven-domain blueprint that adds solution design, governance and risk management, stakeholder communication, and evaluation/optimization, aimed at a more senior audience with a recommended 3+ years in systems architecture and its own 6+ months of Claude-specific experience. Most candidates without prior Anthropic certifications should treat Foundations as the appropriate starting point rather than jumping straight to Professional. See [Claude Certified Architect – Professional](/claude-certified-architect-professional) for that tier's own detail.

All are distinct credentials within Anthropic's four-certification program, each aimed at a different role and technical depth. Claude Certified Architect – Foundations targets solution architects designing agentic systems — orchestration, Claude Code team configuration, MCP tool design, prompt engineering, and context management. Claude Certified Associate and Claude Certified Developer target different audiences and technical scopes within the same overall Claude certification program, with their own separate exam guides, fees, and domain blueprints. Rather than duplicating that detail here, see the [Claude Certifications umbrella page](/claude-certifications) for a full side-by-side comparison of all four credentials, including which one best fits a given role.

Realistically, no — not without first building the same hands-on technical experience the exam expects. CCA-F is an architect-level technical exam that tests applied knowledge of the Claude Agent SDK, Claude Code configuration, MCP tool and resource design, prompt engineering for structured output, and context management. It assumes you've actually built and debugged agentic systems, not just used Claude conversationally or understood its capabilities at a conceptual level. A product manager, sales engineer, or other non-technical stakeholder who understands what Claude can do but hasn't configured Claude Code, built an MCP server, or worked with the Agent SDK directly would be attempting the exam well outside its intended audience, and the scenario-based format would likely expose that gap quickly. If you're in a non-technical role but want to move toward this credential, the honest path is building hands-on technical experience first — see the concrete starting projects on the [eligibility page](/claude-certified-architect/eligibility).

At a high level: first, honestly self-assess against the MQC profile — hands-on experience with Agent SDK orchestration, Claude Code team configuration, MCP tool design, prompt engineering, and context management (the [eligibility page](/claude-certified-architect/eligibility) has a full checklist). Second, close any real gaps with hands-on projects rather than passive reading — shipping a small MCP server or a real CLAUDE.md-driven workflow teaches more per hour than re-reading documentation. Third, review the official exam guide on the Anthropic Partner Academy so you know the exact domain weights and task statements you're accountable for. Fourth, register and complete checkout for the $125 fee, create your Pearson VUE account, and schedule an online-proctored or test-center session, noting the 24-hour cancellation window. Fifth, sit the 120-minute, 60-item exam. Sixth, once you pass, plan your renewal roughly 60-90 days before your 12-month expiry date so you never have to pay for a full retake due to a lapsed credential. The [cost and registration page](/claude-certified-architect/cost-registration) and the [shared policy reference](/claude-certifications/policies) cover the mechanics of steps three through six in full detail.

The authoritative source for exam details, the official exam guide download, and registration is the Anthropic Partner Academy, with exam scheduling and delivery handled by Pearson VUE. Anthropic's own certification program pages are the only place you should trust for the current fee, current domain blueprint, and current policy terms — those details can change between exam guide versions, so a third-party site's summary (including this one) should always be treated as a convenience layer on top of the official source, not a replacement for it. If you're ever unsure whether a page is official, start from the [Claude Certified Architect hub](/claude-certified-architect) or the [exam guide](/exam-guide) on this site, both of which point toward the legitimate registration flow rather than any unofficial reseller.

General Claude API experience helps, but it isn't sufficient on its own for the Tool Design & MCP Integration domain, which accounts for 18% of the scored items — a meaningful share. That domain specifically tests designing MCP tool and resource interfaces: decisions like tool granularity (one broad multi-purpose tool versus several narrow single-purpose tools), how to expose resources versus actions, and how a well-designed MCP server avoids ambiguity that causes an agent to call the wrong tool or misuse the right one. Someone who has only ever consumed existing MCP servers built by others, without having designed one from scratch, typically has a real gap here regardless of how strong their general Claude API experience is elsewhere. If this domain feels abstract to you, building and iterating on even a small custom MCP server is the single highest-leverage prep activity for this specific 18% of the exam.

It's reasonable to expect that it will, given how quickly the Claude Agent SDK, Claude Code, and the Model Context Protocol itself have evolved historically — this is also the core reason Anthropic time-limits the credential to 12 months rather than issuing it permanently (see the [validity page](/claude-certified-architect/validity) for the full reasoning). Practically, this means you should always prepare from the exact exam guide version current at the time you register, rather than from older study notes, blog posts, or forum discussions that may reference a domain blueprint or weighting that's since been revised. It also means third-party prep content — including this site's — has a shelf life, and should be cross-checked against the current official exam guide rather than trusted as permanently accurate.