If you build production systems with Claude, you’ve probably heard the name by now: Claude Certified Architect Foundations. Launched by Anthropic on March 12, 2026, alongside the Claude Partner Network, it’s Anthropic’s first official proctored technical credential, a closed-book, scenario-based exam that validates whether you can make sound architectural decisions with Claude Code, the Agent SDK, the Claude API, and Model Context Protocol (MCP). This guide covers exactly what the certification is, how to register in 2026, the official exam format and passing score, the current domain weights, a realistic study plan, and what it actually does for your career.
What the Certification Actually Is
The Claude Certified Architect Foundations certification (exam code CCA-F, sometimes written CCAR-F or searched as CCAF) is not a course-completion badge you click through in an afternoon. It’s a proctored, scenario-based exam designed to test whether a solutions architect can design and ship production-grade Claude applications at enterprise scale, not recite trivia about model parameters.
According to Anthropic Academy’s official certification page, the exam tests foundational knowledge across four core technology areas: Claude Code, the Claude Agent SDK, the Claude API, and MCP, the technologies most teams rely on to build agentic systems in production. Anthropic has positioned CCA-F as the entry point of a broader, multi-level credential stack that will eventually include tracks for developers, associates, and advanced (Professional-level) architects. You can read Anthropic’s own announcement of the credential and the partner program here: Anthropic Partner Network announcement.
Anthropic has already expanded the program beyond the Foundations level. If you are deciding between the entry-level credential and the more advanced tier, this side-by-side comparison of Claude Certified Architect Foundations vs Professional will help you choose the right exam for your experience level and career stage.
A few things distinguish this claude certified architect foundations certification from typical vendor badges:
- It’s proctored, not self-graded. Sessions run through Pearson VUE, either online with a live proctor or at a physical test center.
- It’s scenario-based, not definition-based. Questions are built around realistic situations, debugging a tool-selection failure, deciding when a support agent should escalate to a human, choosing between a single-agent and hub-and-spoke multi-agent design.
- It launched with real market pull. Major systems integrators and consultancies, including Accenture, Deloitte, and IBM Consulting, moved quickly to treat CCA-F as a credibility signal for client-facing Claude work, and some now require it for architects on Claude engagements.
In short, the Claude Certified Architect Foundations (CCA-F) is Anthropic’s way of formally certifying that an engineer or architect can go beyond “using Claude” to actually architecting reliable, production-grade systems around it.
How to Register Step-by-Step (registration 2026)
Registration 2026 for CCA-F currently runs through a partner-gated access model, which trips up a lot of first-time candidates. Here’s the accurate, current process:
- Confirm your organization is in the Anthropic Partner Network. Eligibility during the current access phase is gated on partner-employer attestation. Check or apply for free via the Anthropic Partner Network announcement.
- Use a corporate email address on a recognized domain. The Skilljar registration flow rejects personal addresses (Gmail, Outlook, Yahoo, iCloud) because they can’t be mapped to a partner employer.
- Request access through the official Skilljar portal. Go to Anthropic Skilljar – CCA-F Access Request and submit the attestation form tied to your work email.
- Complete the free prep courses (optional but recommended). Anthropic Academy hosts 13 self-paced courses that map directly to the exam domains, including “Building with the Claude API,” “Introduction to Model Context Protocol,” and “Claude Code in Action.”
- Schedule your proctored session via Pearson VUE. Once your Skilljar entitlement is approved, you’ll book either an online-proctored session or a seat at a Pearson VUE test center.
- Sit the exam within your entitlement window. Community reporting suggests a roughly 30-day scheduling window from entitlement issuance, though Anthropic hasn’t published a fixed figure, check your Skilljar dashboard for your exact deadline.
- Pay the exam fee (if applicable). The standard fee is $125 USD, but eligible partner-company employees currently receive a 100% discount, bringing early-access seats to $0.
Individuals without a qualifying employer: you can still access the free Anthropic Academy courses today; broader public exam access (without partner gating) is widely expected later in 2026, though Anthropic hasn’t confirmed a date. For the full official record, Exam Guide PDF, Certification Terms and Conditions, and Certification Exam Policy, see the Claude Certified Architect – Foundations page on Anthropic Academy.
Claude Certified Architect Foundations Exam Format & Passing Score
Here’s what the official exam guide and confirmed partner materials say about the claude certified architect foundations exam format:
| Detail | Specification |
|---|---|
| Question count | 60 scenario-based items |
| Question types | Multiple-choice and multiple-response (each item states how many answers to select) |
| Time limit | 120 minutes (no per-question cap) |
| Delivery | Online-proctored or Pearson VUE test center |
| Book policy | Closed-book, no notes, no second monitor, no Claude open during the session |
| Scoring | Scaled score out of 1,000 |
| Passing score | 720 / 1,000 |
| Guessing penalty | None, always submit an answer |
| Certification validity | 12 months from award date (per the current v1.0 exam guide) |
| Exam fee | $125 USD standard; $0 for eligible partner-network employees |
| Result delivery | Typically within a few business days |
A quick note on the passing score: Anthropic uses an equated scaled score rather than publishing a raw “correct answers needed” number, since question difficulty is balanced across exam forms. Treat 720/1,000 as the bar and build in a comfortable margin during practice, most successful candidates report scoring 800+ on realistic mock exams before sitting the real thing.
Proctoring is strict: webcam on for the full session, full desktop screen-share required, no scheduled breaks, and a room sweep that checks for a second monitor or visible phone. Several early candidates lost their session over avoidable setup mistakes, so treat the technical check as seriously as the content review.
Official 2026 Study Guide (domains + weights)
The exam is built around five weighted competency domains. Understanding these weights is the single highest-leverage thing you can do before opening a study guide, because they tell you exactly where to spend your hours.
| Domain | Weight | What it covers |
|---|---|---|
| Agentic Architecture & Orchestration | 27% | Single-agent vs. hub-and-spoke vs. multi-agent design, Agent SDK loop structure, stop_reason handling, subagent context isolation, session state, escalation logic |
| Tool Design & MCP Integration | 18% | Writing deterministic tool descriptions, error-handling contracts, tool_choice modes, MCP servers vs. tools vs. skills |
| Claude Code Configuration & Workflows | 20% | CLAUDE.md hierarchy, Skills frontmatter, Plan Mode, hooks, headless (-p) mode for CI/CD |
| Prompt Engineering & Structured Output | 20% | System-prompt construction, JSON schema output with retry loops, few-shot patterns, instruction hierarchy |
| Context Management & Reliability | 15% | Context-window limits, prompt caching, long-horizon memory strategies, structured human-in-the-loop handoffs |
Domain 1 is by far the largest single domain, roughly one in four questions, and candidates consistently report it as the hardest, since it rewards architectural judgment over memorized syntax. Domain 3 and Domain 4 are tied as the second-heaviest domains, so don’t neglect Claude Code configuration in favor of pure prompting content.
The official exam guide references a scenario pool used to frame questions. Anthropic’s original PDF documented six core scenarios (customer support resolution agents, code generation with Claude Code, multi-agent research systems, developer productivity tooling, Claude Code in CI/CD, and structured data extraction); the live exam has since expanded coverage to include newer surface area like Agent Skills and long-document processing, reflecting product updates shipped after the guide’s original publication.
Sample Question Style
Real exam questions are scenario-driven rather than definition-driven. Here’s the style you should expect (paraphrased for illustration, not lifted from the live exam):
| Sample Question Type | Domain |
|---|---|
| Given a support agent that must decide when to escalate to a human, which architecture minimizes false escalations while preserving safety? | Domain 1 |
| A tool call is failing silently in production. Which of the four isError handling patterns should the architect implement? | Domain 2 |
| Which CLAUDE.md placement takes precedence when both a project-level and user-level file define the same rule? | Domain 3 |
| Which prompt structure most reliably produces valid JSON output on the first attempt without a retry loop? | Domain 4 |
| A long-running agent session is losing earlier instructions. What’s the most effective mitigation? | Domain 5 |
How to Prepare (timeline, resources, checklist)
Most developers with real Claude experience need 15–20 hours of focused study; if you’re newer to the API and Agent SDK, budget 30–40 hours. Here’s a realistic 4-week plan:
Week 1, Foundations & diagnostic
- Take a domain-weighted diagnostic to identify weak areas.
- Work through Anthropic Academy’s “Building with the Claude API” course (the flagship, ~8 hours).
- Review the official exam guide PDF end to end.
Week 2, Deep dive on the heavy domains
- Focus on Domain 1 (Agentic Architecture) and Domain 3 (Claude Code Configuration), together they’re nearly half the exam.
- Complete “Introduction to Model Context Protocol” and “Claude Code in Action.”
- Build one small multi-agent project touching at least three domains.
Reading alone is rarely enough for a scenario-based exam. Work through these practical Claude Architect labs that cover multi-agent systems, MCP integration, Claude Code workflows, and production-style projects mapped directly to the five exam domains.
Week 3, Prompt engineering, structured output, reliability
- Study Domain 4 and Domain 5: system prompt design, JSON schema validation, prompt caching, context-window management.
- Take a full-length timed mock exam under real conditions.
Week 4, Review, retest, book the exam
- Redo weak-domain practice questions.
- Retake a mock exam and confirm you’re consistently scoring above the 720 bar with margin.
- Complete the proctoring tech check and schedule your session.
For a deeper domain-by-domain breakdown, practice questions, and detailed explanations, use this complete Claude Certified Architect Foundations exam prep guide. It expands on every domain weight and includes scenario-style questions that mirror the real exam format.
Free Resources
- Anthropic Academy course catalog, 13 free self-paced courses mapped to exam domains, including Building with the Claude API and Introduction to Model Context Protocol
- Official Claude Certified Architect – Foundations certification page, Exam Guide PDF, Certification Terms and Conditions, and Certification Exam Policy
- Anthropic Partner Network, eligibility and program details
If you prefer structured training over pure self-study, this comparison of the best Claude Certified Architect Foundations course options covers official Anthropic training, instructor-led programs, and practice platforms so you can pick the path that matches your timeline and budget.
Job Titles and Salary Outlook 2026
CCA-F is new enough that Anthropic hasn’t published salary data, and no independent salary survey has enough certified holders yet to produce statistically reliable figures. What we can say with confidence is which roles are actively seeking this credential and how the broader AI-architecture job market has trended.
Holding the CCA-F credential early can meaningfully change how recruiters and consulting firms view your profile. For a full breakdown of job titles, salary ranges, and realistic career paths after certification, see the dedicated Claude Certified Architect Foundations career guide.
Job titles where CCA-F is increasingly listed as a preferred or required qualification:
- AI Solutions Architect
- Claude Integration Engineer
- Agentic Systems Architect
- AI Platform Engineer
- Enterprise AI Consultant (common at Accenture, Deloitte, IBM Consulting, EY, KPMG, PwC)
- Technical Lead, AI Applications
| Role Category | Est. 2025 Salary Range (USD) | Est. 2026 Salary Range (USD) | Est. 2027 Outlook (USD) |
|---|---|---|---|
| AI Solutions Architect (mid-level) | $130,000 – $165,000 | $140,000 – $175,000 | $150,000 – $190,000 |
| Senior AI/Agentic Architect | $170,000 – $210,000 | $180,000 – $225,000 | $195,000 – $245,000 |
| Enterprise AI Consultant | $150,000 – $195,000 | $160,000 – $210,000 | $175,000 – $230,000 |
| AI Platform Engineer | $140,000 – $175,000 | $150,000 – $190,000 | $160,000 – $205,000 |
These are directional, market-informed estimates based on general AI-architecture compensation trends, not figures published by Anthropic or a certification body, treat them as a starting point for your own research, not a guarantee.
The more durable signal is demand: with major consulting firms setting internal targets for hundreds of certified architects each by year-end, holding CCA-F early is functioning as a credibility shortcut in client-facing AI engagements, even before large-scale salary data exists.
Common Pitfalls, Pro Tips & Career Impact
Common pitfalls:
- Registering with a personal email. The Skilljar attestation flow requires a corporate domain, personal addresses are rejected outright.
- Studying only the six scenarios in the original PDF. The live exam draws on a wider scenario pool reflecting product updates shipped after the guide’s initial publication (notably Agent Skills). Don’t treat the PDF as the complete universe.
- Underweighting Domain 1. At 27%, Agentic Architecture is nearly a third of your score, treat it as your primary study focus, not a side topic.
- Assuming all questions are single-answer. The format explicitly includes multiple-response items; always read the instruction line on each question.
- Skipping the tech check. Two monitors, a visible phone, or an open tab have gotten candidates removed from live sessions.
Pro tips:
- Build one real multi-agent project before you sit the exam, hands-on Agent SDK and MCP experience translates directly into scenario-question intuition.
- Practice under timed conditions; at roughly 2 minutes per question, pacing matters more than raw knowledge for many candidates.
- Because there’s no guessing penalty, never leave a question blank.
Career impact: Early candidates who’ve shared feedback after passing consistently describe the domain coverage and official prep materials as genuinely useful preparation. That’s consistent with the broader pattern: early holders report that CCA-F is opening conversations with clients and hiring managers faster than generic “AI experience” claims did, precisely because it’s proctored and scenario-based rather than self-certified. Given that major consultancies are already treating it as a prerequisite for Claude-related engagements, early certification looks like a meaningful differentiator for 2026, even though it’s too new for long-term ROI data. For the most current scenario coverage and any policy updates, always cross-check against the official Exam Guide and Certification Exam Policy rather than third-party recaps.
Free Downloadable Checklist
Before you book your exam, confirm you can check off every item below:
- Employer confirmed in the Anthropic Partner Network
- Registered using a corporate email address
- Completed all 13 Anthropic Academy prep courses (or equivalent hands-on experience)
- Reviewed the official Exam Guide PDF and Certification Exam Policy
- Built at least one project spanning 3+ exam domains
- Scored 800+ on a full-length, domain-weighted mock exam
- Confirmed webcam, single-monitor setup, and quiet room for proctoring
- Scheduled Pearson VUE session (online-proctored or test center)
- Reviewed Domain 1 (Agentic Architecture) one final time, it’s 27% of your score
Ready to get your Claude Certified Architect Foundations certification? Download my free Exam Prep Checklist and start today.
Frequentl Asked Questions FAQ’s
What is Claude Certified Architect Foundations?
It’s Anthropic’s official proctored certification (CCA-F) validating that an engineer can design and ship production-grade applications using Claude Code, the Agent SDK, the Claude API, and MCP.
What is the passing score for the CCA-F exam?
720 out of 1,000 on a scaled score.
How many questions are on the exam, and how long do I have?
60 scenario-based multiple-choice and multiple-response questions, with a 120-minute time limit.
How do I register for CCA-F in 2026?
Through Anthropic’s Skilljar access-request portal, which requires partner-employer attestation and a corporate email address.
How much does the exam cost?
$125 USD standard fee; eligible partner-network employees currently pay $0.
Is the exam open-book?
No. It’s closed-book, requires a webcam and full screen-share, and prohibits having Claude open during the session.
How long is the certification valid?
12 months from the date it’s awarded, per the current exam guide.
Which domain should I study the most?
Agentic Architecture & Orchestration (Domain 1) at 27%, the largest single domain.
Is there a guessing penalty?
No, always submit an answer, since a blank scores the same as a wrong answer.
Is Claude Certified Architect Foundations worth it in 2026?
If you design, build, or operate Claude-powered agents professionally, it’s an early, credible signal, especially given that several major consultancies now treat it as a prerequisite for client-facing Claude work.
