The Claude Certified Architect – Foundations (CCA-F) is Anthropic’s professional certification for practitioners who design, build, and deploy Claude-based systems for businesses. This guide covers what the exam tests, how it is structured, what it costs, and how to prepare effectively.
If you are still evaluating whether to pursue the certification at all, start with the article on what the CCA-F certification covers. If you are ready to book a seat, the CCA-F registration guide covers the registration process step by step.
Exam format
The CCA-F is a timed, proctored assessment administered remotely or at approved testing centers. The exam combines three question types designed to test applied implementation judgment, not just recall.
Question types:
- Multiple choice (single answer): Tests knowledge of specific Claude API behaviors, model configuration parameters, Anthropic’s responsible use policies, and enterprise security requirements. Approximately 35–40% of the exam.
- Multiple choice (multiple answer): Tests understanding of workflow design trade-offs and context architecture decisions where more than one consideration applies. Approximately 20–25% of the exam.
- Scenario-based questions: Presents a real business situation (e.g., a professional services firm with a recurring proposal workflow, a distribution company needing invoice reconciliation automation) and asks the candidate to select the correct implementation approach, identify design errors, or design the appropriate human checkpoint structure. Approximately 35–40% of the exam.
Exam length and timing:
| Parameter | Detail |
|---|---|
| Total questions | 65–75 (varies by exam version) |
| Time limit | 120 minutes |
| Average time per question | ~1.7 minutes |
| Allowed references | None — closed book, no external resources |
| Calculator | Not required |
| Breaks | One optional 10-minute break after question 40 |
The scenario-based questions consume the most time. Candidates who budget 2–3 minutes per scenario question and 60–90 seconds per multiple choice question consistently report finishing with 5–15 minutes to spare for review.
Exam cost
| Item | Cost |
|---|---|
| CCA-F exam registration | $300–$500 (varies by region) |
| Retake fee | Same as initial registration fee |
| Official preparation materials | Included with registration |
| Unofficial third-party prep courses | $100–$300 (not required, not Anthropic-endorsed) |
The $300–$500 range reflects regional pricing differences. Candidates in North America and Western Europe typically see fees toward the higher end of that range. Anthropic partner firms with volume agreements may access preferred pricing for multiple seats.
There is no additional fee for choosing remote proctoring versus in-person testing centers — the registration cost is the same regardless of format.
The exam fee is payable at registration and is non-refundable within 14 days of the exam date. Factor preparation time into your scheduling — a failed first attempt costs the same as the initial registration.
Eligibility requirements
The CCA-F has no formal academic prerequisites. Eligibility is based on demonstrated experience:
Required:
- At least one documented Claude-based implementation for a real business use case (not a tutorial or sandbox project)
- Working familiarity with the Claude API — capable of making API calls, configuring parameters, and handling responses programmatically
- Experience designing at least one context architecture: system prompts, context documents, or instruction sets structured for consistent outputs
Recommended (not required, but strongly correlated with first-attempt pass rates):
- 60–120 hours of total implementation experience or structured exam preparation
- Experience with at least two distinct business workflow types (e.g., a content workflow plus an operations workflow)
- Familiarity with enterprise data handling considerations — even at a basic level
The exam does not require a specific number of years of AI experience. Practitioners with six months of intensive Claude implementation work consistently perform better than those with three years of general AI consulting experience who have not built production Claude systems.
Domain breakdown and study hours
The CCA-F is structured across five domains. The table below shows each domain’s exam weight and the recommended study hours based on what candidates with hands-on Claude experience report needing to feel prepared.
| Domain | Exam weight | Recommended study hours | Primary preparation method |
|---|---|---|---|
| Claude API and model configuration | 20% | 10–20 hours | API documentation review, hands-on API testing |
| Context architecture and AI Foundations design | 25% | 15–30 hours | Build 2–3 context packs for real or simulated businesses |
| Workflow design and business integration | 30% | 20–40 hours | Design and critique 4–6 end-to-end workflow implementations |
| Enterprise security, data privacy, and responsible deployment | 15% | 8–15 hours | Anthropic usage policies, access control design scenarios |
| Adoption tracking and performance measurement | 10% | 5–10 hours | Instrumentation design, acceptance rate tracking methods |
| Total | 100% | 58–115 hours |
Domain 1: Claude API and model configuration (20%)
This domain tests technical proficiency with the Claude API: available models, context window sizes, temperature and top-p configuration, streaming responses, token counting, rate limits, and error handling. Candidates who actively use the API in implementation work typically need 10–15 hours of review here. Those who have primarily used Claude through UI interfaces (Claude.ai rather than the API) should budget 15–20 hours.
Domain 2: Context architecture and AI Foundations design (25%)
The highest-yield preparation domain after workflow design. This tests whether candidates can design the structured context layer — voice guides, client archetypes, decision rules, operating procedures — that makes Claude outputs specific to a business rather than generic. The article on what AI Foundations documents contain is directly relevant here.
Preparation: build two or three complete AI Foundations packages for real or realistic businesses. Review them critically for completeness, specificity, and how they perform when loaded into Claude.
Domain 3: Workflow design and business integration (30%)
The largest domain and the one where most first-attempt failures occur. Scenario-based questions present specific business workflows and ask candidates to identify the correct implementation approach, design appropriate human checkpoints, or diagnose errors in a described implementation.
Preparation: design and document end-to-end implementations for at least four to six distinct business workflow types. For each one, work through: what inputs Claude receives, what the prompt structure looks like, where human review is required, what failure modes exist, and how the workflow connects to adjacent business processes.
Candidates who prepare for Domain 3 by reading about workflow design consistently underperform compared to candidates who prepare by building and critiquing actual implementations. This domain tests judgment developed through practice, not knowledge acquired through study.
Domain 4: Enterprise security, data privacy, and responsible deployment (15%)
Tests knowledge of Anthropic’s acceptable use policies, data handling requirements for enterprise deployments, access control design, and responsible deployment considerations for sensitive use cases. This includes specific scenarios involving regulated industries, PII handling, and situations where Claude should not be the decision-maker.
Preparation: read Anthropic’s usage policy and responsible scaling documentation in full. Work through five to ten access control and data handling scenarios.
Domain 5: Adoption tracking and performance measurement (10%)
The smallest domain by weight but one that distinguishes implementation architects from pure developers. Tests whether candidates know how to instrument a Claude deployment for usage tracking, define and measure acceptance rates, identify adoption failure patterns, and build feedback loops into the system.
Preparation: design an adoption tracking framework for a mid-market company deployment. Define what metrics to track, how to collect them, and what thresholds indicate an adoption problem requiring intervention.
Pass score and what happens if you fail
The CCA-F pass score is a scaled score — Anthropic does not publish the raw percentage required to pass, as the scaling adjusts for exam version difficulty. Based on candidate reporting, most passing scores correspond to approximately 72–75% raw accuracy across all domains.
You receive your results within five business days of completing the exam. The results notification includes:
- Pass or fail status
- Domain-by-domain performance breakdown (not your raw score on each question)
- For failing candidates: a recommended focus area report indicating which domains were weakest
If you fail:
- You must wait 14 days before retaking the exam
- You may take the exam up to three times in a 12-month period
- After three attempts, a 90-day waiting period applies before re-registration
- The retake fee equals the initial registration fee
Domain performance reports from failing attempts are specific enough to redirect preparation meaningfully. Candidates who fail Domain 3 (workflow design) almost always need more hands-on implementation practice, not more study. Candidates who fail Domain 4 (enterprise security) typically need a closer read of Anthropic’s policy documentation, not more implementation experience.
Preparation time: realistic estimates
| Candidate profile | Estimated preparation time |
|---|---|
| Active Claude implementation professional (6+ months building production systems) | 40–60 hours |
| Technical developer with API experience but limited implementation design work | 60–90 hours |
| AI consultant with advisory experience but limited hands-on Claude building | 80–120 hours |
| In-house AI lead or operations manager pursuing certification without developer background | 100–140 hours |
These estimates assume the preparation work is dominated by hands-on practice — building implementations, designing workflows, and reviewing context architectures — not passive reading. Passive study against these time estimates produces pass rates roughly 20–30 percentage points lower than active implementation practice.
Recertification requirements
The CCA-F certification is valid for two years from the date of passing. Recertification requires passing a current version of the exam — there is no abbreviated recertification path.
Anthropic notifies certified architects six months before their recertification deadline. The recertification exam reflects the current Claude API and any new domains added since the original certification version.
Certified practitioners who maintain active Claude implementation work typically find recertification requires 20–40 hours of preparation, as their day-to-day work covers most of the exam content.
How the exam maps to real implementation work
The CCA-F is designed so that candidates who pass it can do the work, not just describe it. The domain structure mirrors what an actual Claude implementation engagement requires:
- Domain 1 (API): The technical foundation every implementation is built on
- Domain 2 (Context architecture): The work that determines whether Claude outputs are useful or generic
- Domain 3 (Workflow design): The work that determines whether the implementation gets adopted or abandoned
- Domain 4 (Security and privacy): The work that determines whether the implementation is safe to deploy at enterprise scale
- Domain 5 (Adoption tracking): The work that determines whether the company can tell if the implementation is working
For companies evaluating implementation partners, this structure is worth understanding. A certified architect has been tested across all five layers of a real implementation. An uncertified developer may have deep expertise in one or two layers and significant gaps in others — typically Domains 2, 4, and 5, which are less visible in technical portfolios but critical to long-term implementation success.
The article on how certified architects differ from non-certified developers covers specific scenarios where these gaps show up in practice. The article on why businesses hire certified Claude architect firms covers what to look for in an implementation partner’s credentials before engaging them.
Frequently asked questions
How many questions are on the CCA-F exam?
The exam contains 65–75 questions depending on the current version. Anthropic updates the exam periodically to reflect changes to the Claude API and implementation landscape.
What is the time limit for the exam?
120 minutes. An optional 10-minute break is available after question 40. The break time is deducted from your total exam time, so use it only if needed.
Can I use external resources during the exam?
No. The CCA-F is a closed-book assessment. No documentation, notes, or external resources are permitted during the exam session.
How long do results take?
Results are delivered within five business days of completing the exam. The results notification includes your pass/fail status and a domain-by-domain performance breakdown.
What is the recommended study sequence for someone starting from scratch?
Domain 1 first (API fundamentals), then Domain 2 (context architecture), then Domain 3 (workflow design). Domains 4 and 5 can be studied in parallel with Domain 3. The workflow design domain should dominate your preparation time, since it carries the most exam weight and requires the most hands-on practice to master.
Is the CCA-F available outside North America?
Yes. Remote proctoring makes the exam available globally. In-person testing centers are concentrated in major metropolitan areas in North America, Western Europe, and select cities in the Asia-Pacific region.
Preparing for the CCA-F or evaluating a certified partner?
The CCA-F exam tests exactly what good Claude implementation requires: context architecture, workflow design, responsible deployment, and adoption tracking — not just API fluency.
Path one: prepare for the exam yourself. Use the domain breakdown above to allocate your 60–120 hours of preparation. Weight heavily toward Domain 3 (workflow design) and Domain 2 (context architecture) — together they account for 55% of the exam and are the domains where preparation time has the highest payoff.
Path two: work with a firm that has already passed. Phos AI Labs holds CCA-F certification and has applied that framework across 400+ AI engagements with companies including Zapier, Coca-Cola, Medtronic, Dataiku, and American Express. The first conversation is a readiness assessment for your business — no deck, thirty minutes. Start here.