Home › Certifications › CCAR-F Certification Guide
Anthropic Certification Guide · 2026CCAR-F Certification Guide 2026: Claude Certified Architect – Foundations
The complete guide to Anthropic’s Claude Certified Architect – Foundations exam — all five domains and their weightings, the six official scenarios, what it costs, the salary hike agentic AI architects command in India and the US, and an 8-week roadmap to pass it.
Authorized Training & Technology Partners
Claude
Claude Certified Architect – Foundations (CCAR-F)
The credential this guide covers
Microsoft
AI-103 — Azure AI Apps & Agents Developer Associate
The other half of the FDE path — read the AI-103 guideClaude Certified Architect – Foundations (CCAR-F) is Anthropic’s official architecture-level credential for building production systems on Claude. It is a 120-minute, proctored, scenario-based exam of 60 items delivered through Pearson VUE, it costs US$125, you need 720 out of 1000 to pass, and the credential is valid for 12 months. It tests five domains: agentic architecture and orchestration, Claude Code configuration and workflows, prompt engineering and structured output, tool design and MCP integration, and context management and reliability.
Unlike a knowledge quiz, CCAR-F is grounded in six realistic production scenarios — from customer support agents to multi-agent research pipelines and CI/CD integrations. You are not asked what a feature is called. You are asked which architecture is correct for a described system, and which common approach would quietly break it.
You will see this credential written several ways — CCAR-F, CCA-F and CCFA all appear across the web and in course listings. They all refer to the same Anthropic Claude Certified Architect – Foundations certification. We use CCAR-F throughout this guide, which is the code Anthropic uses.
CCAR-F exam at a glance
| Attribute | Detail |
|---|---|
| Credential | Claude Certified Architect – Foundations (CCAR-F) |
| Issued by | Anthropic |
| Level | Architecture — the practitioner tier above associate and developer |
| Format | 60 scenario-based items — multiple-choice and multiple-response, each stating how many answers to select |
| Duration | 120 minutes, proctored |
| Scoring | Scaled 100–1000; 720 required to pass |
| Scenarios | Drawn from a published bank of 6; 4 are presented at random in any given sitting |
| Exam fee | US$125 (checkout price reflects any applicable partner-tier discount) |
| Delivery | Pearson VUE — online proctored or at a test centre |
| Registration | Anthropic Partner Academy, tied to the Claude Partner Network (free to join at Registered level) |
| Validity | 12 months from the date awarded |
| Prerequisites | None enforced — but real Claude build experience is assumed throughout |
What is the Claude Certified Architect credential?
CCAR-F is part of Anthropic’s first professional certification family, and it sits at the architecture level. Where an associate-tier credential asks whether you can use Claude, CCAR-F asks whether you can design a Claude system that keeps working in production — under real traffic, real tool failures, real context pressure and real compliance constraints.
That framing explains the exam’s unusual shape. Every item is anchored to a scenario. The wrong answers are not nonsense; they are the plausible, commonly-shipped approaches that fail at scale. If you have built agentic systems and watched them break, the exam reads as familiar. If you have only read about them, it reads as a series of near-identical options.
CCAR-F exam domains and weightings
Five domains, unevenly weighted. Agentic architecture alone is over a quarter of the exam — nearly double the smallest domain. Your study plan should mirror these proportions rather than treating the five as equals.
CCAR-F exam blueprint — share of exam by domain
Five domains, 60 scenario-based items, 120 minutes
Bars scaled to the largest domain. Weightings as published for the Claude Certified Architect – Foundations exam, 2026.
Agentic Architecture & Orchestration
The agentic loop lifecycle: stop_reason handling, tool_use versus end_turn, keeping tool results in history. Model-driven decisions versus pre-configured decision trees. Hub-and-spoke coordinator–subagent design. Explicit context passing — subagents inherit nothing. Punished anti-patterns: parsing text to decide termination, and arbitrary iteration caps.
Claude Code Configuration & Workflows
CLAUDE.md hierarchy across user, project and directory levels; imports and memory diagnostics. Rules with glob frontmatter. Commands and skills. Plan mode versus direct execution — complexity assessment is tested hard. Headless CI/CD: print mode, JSON output, schema-constrained results, independent-instance review.
20% of examPrompt Engineering & Structured Output
Designing prompts whose output is reliably machine-parseable, using schema constraints rather than hopeful formatting instructions. Producing results downstream code can parse deterministically — and handling the cases where it cannot.
20% of examTool Design & MCP Integration
Description engineering that never misroutes: inputs, edge cases, boundaries against similar tools, when to split versus consolidate. Structured errors with isError, categories and retryable flags. Why an oversized tool set degrades selection. .mcp.json project versus user scope, and tool_choice modes.
Context Management & Reliability
Context window strategy on long-running work. Session resume versus fork versus fresh-with-summary. The reliability patterns that keep multi-step systems correct as individual steps degrade — the failure modes that only appear after hours of running.
15% of examHow the exam actually catches you
Distractors are not nonsense — they are the approaches most teams ship. Knowing why an approach fails is worth as much as knowing what works. Study the anti-patterns as hard as the patterns.
Applies to all 5 domainsThe six official scenarios
This is the structural detail that makes CCAR-F preparable in a way most exams are not. The scenario bank is published, and a sitting draws four of the six at random. Build all six before you sit, and you cannot be surprised by the context — only by the question.
Customer support resolution agent
An agent with MCP tools resolving real support cases — loop control, tool routing and safe termination under live traffic.
Domain 1 & 2Team Claude Code configuration
Configuration hierarchy, rules and skills for a working team repository, with the plan-mode judgement calls that separate a usable setup from a noisy one.
Domain 3Multi-agent research pipeline
Coordinator and subagents with provenance tracking — explicit context passing, parallel spawning, and the “too narrow” decomposition failure.
Domain 1Tool suite & MCP integration layer
A tool set that routes ambiguous requests correctly, with structured, recoverable errors and scoping that survives a growing catalogue.
Domain 2CI/CD PR-review pipeline
Headless Claude Code producing structured inline findings, with false-positive control and prior-findings context to avoid duplicate comments.
Domain 3Structured output & context reliability
Schema-constrained generation and context strategy on long-running production work, where the failure modes only surface at scale.
Domain 4 & 5Is the CCAR-F certification useful?
We are going to be straight with you here, because most content on this topic is not.
There is no credible published figure for the salary uplift attributable to a Claude certification specifically. Any article quoting one is guessing. Claude credentials are new and do not yet carry the HR familiarity that AWS and Microsoft certifications accumulated over a decade. What can be said accurately is where the credential does carry weight, and what the underlying skills pay — both covered below.
Where CCAR-F genuinely carries weight
- Claude Partner Network firms. The strongest signal by far. Consulting and services firms in the partner network — several with large Indian delivery operations — have direct commercial reasons to value certified architects, since partner tier and staffing credibility depend on it.
- Being early. The credential is new. Being among the first certified architects in your market is a positioning advantage that will not exist in three years.
- The curriculum itself. Agentic loop correctness, MCP tool design and context strategy are exactly the skills that separate an agent demo from an agent system. Very little of this is taught coherently anywhere else.
- Pairing with a hyperscaler credential. CCAR-F plus a Microsoft or AWS AI credential covers both the model layer and the infrastructure layer — a materially stronger profile than either alone.
Where it will not carry you
- Generic HR screens. A recruiter outside the AI ecosystem may not recognise the credential yet. Your deployed work still does the heavy lifting.
- As a substitute for engineering ability. The exam assumes you can build. It certifies judgement, not raw capability.
- Indefinitely. The credential is valid for 12 months, so it is a recurring commitment.
What agentic AI architects actually earn
Since no figure attaches to the certificate itself, the useful question is what the role pays — agentic AI engineer, AI architect, LLM systems engineer. That market data does exist, and it is unusually strong.
What the hike looks like in rupees
The reported agentic premium, applied at its midpoint (~40%) to published Indian software engineering bands. This is the gap the CCAR-F curriculum is built to close.
Entry · 0–2 yrs
Mid · 3–5 yrs
Senior · 6–9 yrs
Architect · 10+ yrs
Agentic AI and architect pay by market
India and the US are shown as two separate charts because rupee and dollar bands are different measures on different scales; forcing them onto one axis would distort both.
India — annual CTC
Agentic AI / LLM architect roles, ₹ lakh per annum
Bar length maps the upper bound of each band against a ₹52 L scale. Sources: Glassdoor India AI Architect/Engineer data (July 2026, avg ~₹35 L, typical range ~₹20.75–47.75 L); reported ₹20–40 L+ for engineers with RAG, agentic workflow and system-design skills.
United States — annual base
Claude / LLM engineering roles, US$ thousands
Bar length maps each figure against a $200K scale. Sources: ZipRecruiter Claude AI job postings ($57K–$160K, May 2026); published prompt-engineering ranges of $60K–$250K+ in the US.
Generalist vs agentic specialist
Generalist software engineer vs agentic AI engineer
India, ₹ lakh per annum — modelled from the reported 30–50% agentic premium
Method: generalist bands drawn from published Indian software engineering ranges; the specialist bar applies the reported 30–50% agentic premium at its midpoint (~40%). Illustrative modelling, not survey data — treat it as direction, not a quotation.
Compensation figures are compiled from independent, publicly available industry sources and are shown for role context. They are not a guarantee of pay in any specific market, company or outcome, and 360DT does not promise a salary result from any certification or programme.
Why an architecture credential for Claude exists
1. LLM expertise is the premium skill category of 2026
Reporting through 2026 consistently places large language model expertise — training, fine-tuning and deploying models including Claude — at the top of the compensation stack. Agentic AI engineering is described as the successor to the 2023 “LLM engineer” wave, and a better-defined one: firmer technical scope around orchestration frameworks, MCP and multi-agent evaluation.
2. The failure mode moved from capability to architecture
Two years ago, AI projects failed because models could not do the task. They now typically fail on architecture: agents that terminate on the wrong signal, tool sets so large the model misroutes, subagents starved of context, outputs that parse correctly nine times in ten. Every one of those is a named CCAR-F objective. The exam exists because the industry’s bottleneck moved.
3. Partner-network economics
Registration runs through the Anthropic Partner Academy, tied to the Claude Partner Network, and the exam fee reflects partner-tier discounts. That structure tells you who the credential is built for: services and consulting firms deploying Claude for clients, who need a defensible way to demonstrate staffed capability. For engineers in or near those firms, this is the credential with commercial pull.
Treat CCAR-F as a skills-forcing function with a partner-market signal attached, not as a salary lever on its own. Sit it because the five domains are the real production curriculum for agentic systems — and because being early in a new credential is worth more now than it will be later.
Your 8-week CCAR-F roadmap
Weighted to the blueprint, structured around building all six scenarios, and reserving the final week for timed, domain-reviewed mocks. Assume 8–10 focused hours per week.
Agentic loops done correctly
The loop lifecycle: stop_reason handling, tool_use versus end_turn, keeping tool results in history. Model-driven decisions versus pre-configured decision trees. Learn the two punished anti-patterns cold: parsing text for termination, and arbitrary iteration caps. Begin Scenario 1.
Multi-agent systems, hooks and enforcement
Hub-and-spoke coordination and task decomposition pitfalls. Explicit context passing — subagents inherit nothing. Parallel spawning. PostToolUse hooks for normalisation and interception hooks for compliance limits. Session resume versus fork versus fresh-with-summary. Build Scenario 3.
Tool design and MCP integration
Description engineering that never misroutes. Structured errors with categories and retryable flags; access failure versus valid-empty. Why an oversized tool set breaks selection; scoped cross-role tools; tool_choice modes. .mcp.json project versus user scope. Build Scenario 4.
Claude Code configuration and workflows
CLAUDE.md hierarchy and imports, rules with glob frontmatter, commands and skills. Plan mode versus direct execution — complexity assessment is tested hard. Iterative refinement with I/O examples and test-driven iteration. Build Scenario 2.
Claude Code in CI/CD
Headless print mode, JSON output formats and schema-constrained results in pipelines. Independent-instance review versus self-review; passing prior findings to avoid duplicates. CLAUDE.md as CI context. Build Scenario 5.
Prompt engineering and structured output
Schema-constrained generation rather than formatting instructions and hope. Designing outputs that downstream code can parse deterministically, and handling the cases where they cannot. Build Scenario 6.
Context management and reliability
Context window strategy on long-running work, session lifecycle decisions, and the reliability patterns that keep multi-step systems correct as individual steps degrade. Then review all six scenario builds end to end.
Timed mocks and the exam
Full 120-minute, 60-item mocks weighted to the real blueprint, reviewed domain by domain. Any domain under 75% gets a dedicated re-study day. Complete Pearson VUE registration through the Anthropic Partner Academy and book the slot.
How to prepare for CCAR-F
- Build all six scenarios. Four appear at random. Building all six removes the luck.
- Study the anti-patterns as hard as the patterns. Distractor answers are the plausible wrong approaches, so knowing why something fails is worth as much as knowing what works.
- Weight your time 27/20/20/18/15. Even coverage will underprepare you on agentic architecture, the largest domain.
- Practise reading multi-response items. Each item states how many answers to select; misreading that instruction under time pressure costs marks you had earned.
- Join the Claude Partner Network early. Registration runs through the Partner Academy and is free at the Registered level; sort it out before exam week, not during it.
Learn from Anthropic Authorized Instructors
360DT delivers Claude certification prep live, taught by Anthropic Authorized Instructors alongside Microsoft, AWS, NVIDIA and Cisco certified trainers — our full international faculty, and the same people who run our Forward Deployed Engineer cohorts.
Vikas Mittal
Google and Anthropic Authorized AI Instructor
Veteran technology trainer with more than 25 years of rich industry experience spanning Google, Microsoft, and emerging AI platforms. An exceptional mentor recognized for proven teaching excellence.
Deep
Anthropic Authorized Instructor and AI Cloud Specialist
Seasoned technology leader with over 20 years of experience across AI, cloud computing, and enterprise training. Known for simplifying complex ideas and delivering practical, real world learning.
Sid
Anthropic Authorized Instructor and Azure Data Engineer
Dynamic corporate trainer and consultant with more than 10 years of experience across Microsoft Fabric, Azure Data Engineering, and modern cloud platforms. Focused on clear, practical learning that sticks.
Bipeen
AWS, Microsoft, and Anthropic Certified Trainer
AI, machine learning, and cloud transformation expert with more than 25 years of experience. A renowned global corporate trainer and conference speaker known for practical, outcome driven sessions.
Shantanu Pandey
Microsoft, Google, NVIDIA and HPE Certified Trainer
Director of Engineering at Meteoros Automation, ranked in Microsoft's Top 100 Trainers of 2025, holding MCT, Google Cloud, NVIDIA and PeopleCert credentials, with 500+ trainings delivered since 2011.
Arshad Ahmad
Microsoft Certified Trainer (MCT), Cybersecurity & Power Platform Expert
Technology Trainer with 15 years in strategic client delivery, holding MCT, Azure Solutions Architect Expert, Cybersecurity Architect Expert and Power BI Data Analyst credentials, with 200+ trainings delivered.
Sravia
NVIDIA Authorized Instructor and AI Data Specialist
Passionate AI and data evangelist with over 10 years of experience across machine learning, MLOps, and cloud native technologies. Dedicated to delivering impactful, engaging enterprise training programs.
Johan
Microsoft Business Applications MVP and Microsoft Certified Trainer
D365 Customer Service Lead – Europe at Avanade, and founder of the Power Platform School. A Microsoft Business Applications MVP and Microsoft Certified Trainer based in London, specializing in Dynamics 365 Customer Service and Power Platform.
Akim Nyamande
Microsoft, CompTIA, Juniper and Cisco Certified Trainer
IT training facilitator with over five years of experience and certifications across Microsoft, CompTIA, Juniper, and Cisco. Previously a Network Administrator before moving into technical training, now delivering hands-on courses in networking, systems administration, and cybersecurity fundamentals.
Ali El Khatib
Cisco Certified Systems Instructor (CCSI) and Microsoft Certified Trainer
Infrastructure Engineer at RHUH with 13 years in the training field, holding CCSI, MCSE, CCNP Routing and Switching, CCNP Security, CompTIA and Microsoft Azure certifications.
Florian Garcia Compte
Cisco Certified Systems Instructor (CCSI)
Partner Director at AVAANZA FORMACION, a Cisco Learning Partner based in Madrid. Certified Cisco Systems Instructor (CCSI No. 21053) specializing in Cisco Data Center (ACI/SDN), Nexus, and Wireless, delivering official Cisco certification courses including CCNA, CCNP Enterprise, and CCNP Data Center.
CCAR-F is one half of the Forward Deployed Engineer path
The 360DT Forward Deployed Engineer Program covers Claude Certified Architect – Foundations and Microsoft’s AI-103, then goes further: RAG, agents, MCP, deployment and the client-facing delivery skills that turn a certified architect into the engineer AI labs compete to hire.
Explore the FDE Program
CCAR-F frequently asked questions
How much does the CCAR-F exam cost?
US$125. The price shown at checkout reflects any discount that applies to your Claude Partner Network tier. The exam is delivered by Pearson VUE, online proctored or at a test centre, and the fee is paid directly to the exam provider rather than to a training company.
Will CCAR-F increase my salary?
No credible published figure attaches a salary uplift to a Claude certification specifically, and anyone quoting one is guessing. What is documented is that the underlying role — agentic AI engineering — carries a reported 30–50% premium over generalist software engineering, widening to 2–3× for senior multi-agent talent. Applied to Indian bands at the midpoint, that is roughly +₹2.5 L at entry, +₹5 L at mid, +₹9 L at senior and +₹13 L at architect level. The credential is most commercially useful inside Claude Partner Network firms, several of which have large Indian delivery operations.
What is the CCAR-F passing score?
720 out of 1000 on a scaled 100–1000 range. The exam is 60 scenario-based items in 120 minutes, mixing multiple-choice and multiple-response questions — each item tells you how many responses to select.
Is CCAR-F the same as CCA-F and CCFA?
Yes. All three refer to the same Anthropic Claude Certified Architect – Foundations certification. CCAR-F is the code Anthropic itself uses and the one we use throughout this guide; CCA-F and CCFA appear across training providers, course listings and community sites.
How long is the CCAR-F credential valid?
12 months from the date it is awarded. Plan for re-certification as a recurring commitment rather than a one-time achievement.
Do I need to join the Claude Partner Network to sit the exam?
Registration runs through the Anthropic Partner Academy, which is tied to the Claude Partner Network. Joining starts free at the Registered level, so this is an administrative step rather than a commercial barrier — but complete it well before you intend to book.
How long does it take to prepare for CCAR-F?
Roughly 8 weeks at 8–10 focused hours per week for an engineer with existing Claude and agent-building experience — the roadmap on this page. Candidates who have not built agentic systems before should expect longer, since the exam tests judgement earned by building rather than material that can be memorised.
Which is better, CCAR-F or a Microsoft AI certification?
They are not competitors — they test different layers. CCAR-F covers the model layer: agentic architecture, tool and MCP design, context reliability. Microsoft’s AI-103 covers the infrastructure layer: deployment, grounding, security, cost and monitoring on Azure. Holding both is what a Forward Deployed Engineer role genuinely asks for, which is why our FDE programme prepares for both.
Sources and further reading
- Anthropic — Claude Certified Architect – Foundations credential, delivered via Pearson VUE and registered through the Anthropic Partner Academy
- Published CCAR-F exam specification — 60 scenario-based items, 120 minutes, scaled 100–1000 with a 720 pass mark, US$125, 12-month validity
- Published CCAR-F domain weightings — Agentic Architecture & Orchestration 27%, Claude Code Configuration & Workflows 20%, Prompt Engineering & Structured Output 20%, Tool Design & MCP Integration 18%, Context Management & Reliability 15%
- ZipRecruiter (May 2026) — Claude AI job postings, $57K–$160K
- Glassdoor India (July 2026) — AI Architect/Engineer average ~₹35 L, typical range ~₹20.75–47.75 L
- Independent 2026 industry reporting on the agentic AI engineering premium (30–50% over generalist software engineering, widening to 2–3× at senior multi-agent level) and on the absence of a credible Claude-certification-specific salary figure
360DT is an independent training provider. Certification exams are administered independently by Anthropic through Pearson VUE and are not included in programme tuition. Exam details are accurate as of 20 August 2026; always confirm current format, pricing and domain weightings with Anthropic before booking.