Home › Certifications › Claude Architect for Agentic AI Engineers
Agentic AI Certification Guide · 2026Claude Certified Architect – Foundations for Agentic AI Engineers: The 2026 Guide
The architecture half of an agentic AI engineering skillset. Why 45% of CCAR-F is agent orchestration and tool design — the highest agentic concentration of any credential we cover — what MCP actually changes, how Claude Code fits, why context reliability gets its own domain, and a 12-week roadmap that pairs with AI-103.
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 platform-side half of the agentic stack — read the companion guideClaude Certified Architect – Foundations (CCAR-F) is Anthropic’s architecture-tier credential, and it is the most agent-dense certification in this series. 60 scenario-based questions, 120 minutes, 720 out of 1000 to pass. Agentic Architecture & Orchestration is 27% and Tool Design & MCP Integration is 18% — 45% of the exam on how agent systems should be shaped, before a single line of platform-specific code.
This guide reads CCAR-F specifically as the architecture half of an agentic AI engineer’s toolkit, paired with Microsoft’s AI-103 on the platform side. For the general walkthrough — every domain, the six-scenario bank, registration through the Partner Academy — see our complete CCAR-F certification guide, which frames the same exam as one half of the Forward Deployed Engineer path.
Why agentic AI needed an architecture credential
Building one agent is a weekend. Building an agent system that a business depends on is a different discipline, and the failures are architectural rather than syntactic: an agent that loops forever because nobody defined a stop condition, a tool suite so overlapping the model cannot choose between them, context that degrades silently at length, a multi-agent design where three agents do the work one would have done better.
None of those are coding mistakes. They are design mistakes, and they are what a scenario-based architecture exam is built to test. That is the gap CCAR-F sits in — and why its questions read as situations rather than syntax.
CCAR-F domains and weightings
Five domains, with exact percentages rather than ranges. Read them as a statement about what Anthropic thinks matters.
CCAR-F exam blueprint — share of content by domain
Official Anthropic weightings, Claude Certified Architect – Foundations
Source: Anthropic Partner Academy, CCAR-F exam blueprint. Percentages total exactly 100%. Applied to 60 questions: roughly 16 orchestration, 12 Claude Code, 12 prompting, 11 tool and MCP, and 9 context-reliability items.
Agentic Architecture & Orchestration (27%) plus Tool Design & MCP Integration (18%) means nearly half this exam is about how agents are structured and how they reach the outside world. Compare that with AI-103, where the agentic domain is 30–35%. Both are agent credentials; only one puts agent design at its centre.
Where each half of the stack carries its weight
The same comparison from the other side. This is the clearest single reason to hold both rather than choosing.
Share of each exam given to agentic architecture and tooling
CCAR-F and AI-103, by published blueprint weighting
Calculated by us from each vendor’s published blueprint. CCAR-F: Agentic Architecture & Orchestration 27% plus Tool Design & MCP Integration 18% = 45%. AI-103: Implement generative AI and agentic solutions, published as a 30–35% range, shown at its 32.5% midpoint. The two are complementary rather than competing — CCAR-F carries the design weight, AI-103 carries the platform weight.
The five domains as flashcards
Agentic Architecture & Orchestration
When a single agent is right and when a multi-agent design is genuinely better rather than merely fashionable. Orchestration topologies, delegation and hand-off between agents, bounded loops and stop conditions, and failure handling when a step goes wrong halfway through. The most common correct answer is the simpler architecture — and recognising that is the skill being tested.
27% · largest domainClaude Code Configuration & Workflows
Configuring Claude Code for real engineering work: project setup, permissions and what an agent is allowed to touch, workflow design, and integrating it into an existing development process. A fifth of the exam on a working tool rather than on theory — which is unusual, and tells you Anthropic expects candidates to have actually used it.
20% of the examPrompt Engineering & Structured Output
Prompting as an engineering practice, not a party trick: system design, few-shot construction, and above all structured output that downstream code can parse without defensive string handling. In an agent, the model’s output is another component’s input — reliability here is what stops the whole chain from being brittle.
20% of the examTool Design & MCP Integration
Designing a tool suite a model can navigate: clear boundaries, unambiguous descriptions, sensible parameters, no two tools that overlap. Then MCP — the Model Context Protocol — as the standard interface between models and the tools and data they use. Write the integration once and it works across models, which is why this domain exists at all.
18% · the MCP domainContext Management & Reliability
What stays in the window, what gets summarised, what is stored and retrieved. Degradation over long runs and how to detect it before a user does. The smallest domain at 15% — and the one that most separates people who have run an agent in production from people who have demoed one.
15% · smallest domainFour of six scenarios, drawn at random
CCAR-F scenarios come from a published bank of six, of which four appear in any given sitting. Anthropic publishing the bank is a deliberate choice: it moves the exam away from recall and toward whether you can reason inside a situation you have already read. Study all six. You do not get to choose which four you meet.
Published scenario bankCCAR-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 |
| Pairs with | AI-103, Microsoft’s Azure AI Apps and Agents Developer Associate |
Who this pairing is for
- Engineers who have already shipped something with Claude and keep hitting the same architectural walls — loops that will not terminate cleanly, tool suites the model misuses, context that quietly degrades.
- Anyone building on MCP, or about to. 18% of the exam is tool design and MCP integration, and it is the closest thing to a formal credential on the protocol.
- Platform-certified engineers who want vendor-neutral judgement. If you hold AI-103, this is the credential that stops you designing every problem as an Azure problem.
- Technical leads who review other people’s agent designs — the domain weightings map almost exactly onto what goes wrong in an agent design review.
Where it will not carry you
- It is Foundations tier. Above it sits CCAR-P, the professional-tier architect credential, for genuinely senior scope.
- It will not teach you cloud infrastructure. Deployment, scaling, monitoring and cost are platform concerns, and that is exactly what AI-103 covers on the other side.
- It expects you to have built. No prerequisite is enforced, but scenario-based questions are unforgiving of candidates who have only read about agents.
What an agentic AI architect earns
Agent architecture is a scarce skill attached to an abundant job title. Plenty of people can call a model; far fewer can design a system of agents that stays up — and the gap between those two groups is where the upper bands sit.
The AI architect progression
Entry
Mid
Senior
Architect
AI architect and agentic specialist pay by market
India and the US shown as two separate charts, because rupee and dollar bands are different measures on different scales and should never share an axis.
India — annual CTC
AI engineer and architect roles, ₹ lakh per annum
Bar length maps the upper bound of each band against a ₹53 L scale. Bands compiled from published Indian AI engineer and AI architect ranges. Product companies and global capability centres sit at the top of each band.
United States — annual base
AI engineer and agentic specialist roles, US$ thousands
Bar length maps each figure against a $200K scale. Bands compiled from published US AI engineer and prompt/agentic specialist ranges; base pay only, excluding bonus and equity. The spread here is unusually wide because the specialism is young and titles are inconsistent.
These are role bands, not certification outcomes. No published figure attaches a salary increase to holding CCAR-F alone — the credential is too new for that data to exist, and anyone quoting one is guessing. What is defensible: it is an architecture-tier credential from the lab that builds the model, and there are very few people holding it. Scarcity is the honest argument, not a salary multiple.
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 CCAR-F is the architecture half worth having
1. It is vendor-neutral where it counts
CCAR-F tests how agent systems should be designed, and that reasoning survives a change of cloud. An orchestration topology that is right on Claude is right on any capable model; a bad tool schema is bad everywhere. That is the opposite of a platform certification’s shelf life, and it is why pairing one of each is stronger than holding two of either.
2. MCP is becoming the integration standard
The Model Context Protocol lets you write a tool or data integration once and expose it to any model that speaks it, instead of rebuilding per vendor. An 18% domain on tool design and MCP integration is, in effect, a formal credential on the emerging standard — and there are not many of those yet.
3. Claude Code is examined as a working tool
Twenty percent of the exam on configuring and running Claude Code inside a real development workflow is a strong signal about what Anthropic expects an architect to be doing day to day. It is also the domain you cannot fake: candidates who have used it recognise the questions immediately, and candidates who have not, do not.
4. Twelve-month validity keeps it honest
CCAR-F is valid for 12 months. In a field where the tooling is rewritten annually that is appropriate rather than punitive — a current holder has been tested on the current practice. Budget for the renewal as part of the cost, exactly as you would for a Microsoft associate credential.
Your 12-week CCAR-F roadmap
Weighted to the blueprint, with the two agentic domains leading because they are 45% of the marks. Assume 6–8 hours per week, and build with Claude throughout — scenario questions punish reading-only preparation harder than most exams.
Read all six scenarios first
Anthropic publishes the scenario bank; four of the six appear in your sitting. Read every one before you study anything else, so the whole syllabus lands as an answer to a problem you have already met rather than as abstract material.
Single-agent architecture
The plan-act-observe loop, stop conditions and step limits, error handling mid-run. Build one from scratch. Then make it fail on purpose and fix each failure — that catalogue of failures is what the orchestration questions are drawn from.
Multi-agent orchestration
Topologies: supervisor, pipeline, peer delegation. Hand-off and shared state. And the question that carries the most marks: when is one agent simply better? Splitting work across agents adds coordination cost, and the exam rewards candidates who notice that.
Orchestration under failure
Partial completion, retries and idempotency, timeouts, and what a half-finished agent run leaves behind. Observability for a non-deterministic system. Completes the 27% domain — the largest on the exam, banked by week four.
Tool design
Boundaries, naming, descriptions and parameter design. Build a suite of six tools, then deliberately make two of them overlap and watch the model hesitate. Tool design is prompt engineering wearing a schema, and this week is where that clicks.
MCP end to end
The protocol itself: servers, transports, capabilities. Build one MCP server exposing real tools and connect it. Then connect the same server to a second client and see what “write once” actually buys you. Completes the 18% domain.
Prompt engineering as engineering
System prompt design, few-shot construction, decomposition. Prompts under version control and tested like code, because in an agent a prompt regression is a production incident.
Structured output you can rely on
Schema-constrained output, validation and what to do when it still comes back malformed. Every downstream component depends on this holding. Completes the 20% prompting domain.
Claude Code, properly configured
Project configuration, permissions and boundaries on what the agent may touch, workflow design, and integration into an existing development process. Use it daily this week; 20% of the exam assumes you have.
Context management and reliability
Window strategy, summarisation, memory stores and retrieval. Long-run degradation and how to detect it. Reliability patterns for a system that is non-deterministic by construction. Completes the final domain.
Re-read the scenarios, now as an architect
Return to all six with the full blueprint behind you. For each, write down the architecture you would propose and the trade-off you would defend. That exercise is the exam, in a lower-pressure form.
Timed practice and the exam
Full 60-question, 120-minute sittings scored by domain. Rehearse the multiple-response format — each item states how many answers to select, and misreading that under time pressure costs marks you had earned. Then book.
CCAR-F is scenario-based, which means the constraint usually arrives before the question. Identify what is being optimised for — latency, cost, reliability, operational simplicity — before you look at the options. Several answers will be workable architectures; only one serves the stated constraint. And when two designs both fit, the simpler one is usually correct.
Learn from Instructor at 360DT, an Anthropic Partners
Certification prep at 360DT is delivered live by AWS, Microsoft, Cisco and NVIDIA certified trainers and Instructor at 360DT, an Anthropic Partners — our full international faculty.
Naga Nandhini
Java Full Stack Corporate Trainer and Enterprise Solutions Expert
Powerhouse corporate trainer with 10+ years of upskilling engineers at HCL, Accenture, Cybage, Mobileum and Renault. An award-winning engineer who helped build the RAPTOR and KIKO platforms for an American Airlines client, she brings real enterprise delivery into every session.
Praveen Kumar
Full Stack and GenAI Corporate Trainer, Ex-TCS Team Lead
Full stack lead and GenAI builder with 10+ years of training behind 50+ corporate batches for TCS, HCL, Perficient, Renault Nissan and Saint-Gobain. He ships real RAG and local-LLM systems with Ollama, FastAPI and Spring Boot, then teaches exactly how they work.
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.
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.
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.
Vikas Mittal
AI instructor with 25+ years in technology training
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.
Sid
Microsoft Certified Trainer since 2016 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.
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.
The Agentic AI Engineer Program
The 360DT AI Engineer Program prepares you for Claude Certified Architect – Foundations and Microsoft’s AI-103, with the exam blueprints mapped directly into the curriculum. Along the way you build production agents with LangChain, LangGraph, CrewAI and MCP, ship real RAG pipelines with vector search and tool calling, and finish with eight capstone projects rather than a certificate alone.
Explore the AI Engineer Program
Agentic AI and CCAR-F frequently asked questions
How much of CCAR-F is actually about agents?
45% — Agentic Architecture & Orchestration at 27% plus Tool Design & MCP Integration at 18%. Add Context Management & Reliability at 15% and the case is stronger still, since context strategy is an agent concern above all. It is the most agent-dense credential in this guide series.
How is this different from your main CCAR-F guide?
This one reads CCAR-F as the architecture half of an agentic AI engineering skillset, paired with AI-103 on the platform side. Our complete CCAR-F certification guide is the general walkthrough — every domain in depth, the Partner Academy registration route, how the tier sits against CCAO-F and CCAR-P — and frames the same exam as one half of the Forward Deployed Engineer path.
Do I need to be a Claude Partner to sit the exam?
Registration runs through the Anthropic Partner Academy, which is tied to the Claude Partner Network — and joining at the Registered level is free. The US$125 checkout price reflects any partner-tier discount that applies to you.
What is MCP, in one paragraph?
The Model Context Protocol is an open standard for connecting models to tools and data sources. Instead of writing a bespoke integration for every model you use, you expose a capability once through MCP and any client that speaks the protocol can use it. On CCAR-F it shares an 18% domain with tool design, because designing the tool well and exposing it correctly are the same job.
Is Claude Code really 20% of the exam?
Yes — Claude Code Configuration & Workflows is a fifth of the blueprint. That is unusual for a certification and quite deliberate: Anthropic is examining whether you can put an agent to work inside a real development process, not just describe one. The practical implication is simple — use Claude Code for actual work before you sit the exam.
Should I take CCAR-F or CCAR-P?
Foundations first unless you are already architecting Claude systems at scale. CCAR-P is the professional tier above this one and assumes the Foundations material rather than re-teaching it. If you are pairing with AI-103 as an agentic engineer, CCAR-F is the right level.
Which should I take first, CCAR-F or AI-103?
AI-103 first, for most people. Building agents on a concrete platform gives you the experience that makes CCAR-F’s scenarios read as decisions you have faced rather than abstractions. If you have already been building with Claude in production, going straight to CCAR-F is perfectly reasonable.
Are the exams included in the programme fee?
No. Certification exams are administered independently by Anthropic and Microsoft and are not included in programme tuition. The programme fee of ₹27,999 covers the live training, projects and certification preparation; you book and pay for each exam directly with its issuer.
Sources and further reading
- Anthropic Partner Academy — Claude Certified Architect – Foundations exam blueprint (five domains with exact weightings, 60 scenario-based items, 120 minutes, scaled 100–1000 with a 720 pass mark, the six-scenario bank with four presented per sitting, US$125 fee, 12-month validity)
- Anthropic — Claude Partner Network registration tiers, used for the free Registered-level route into the Partner Academy
- Anthropic — Model Context Protocol documentation, used for the MCP description
- Microsoft Learn — AI-103 skills measured, used for the agentic-weight comparison (Implement generative AI and agentic solutions, 30–35%)
- Agentic-weight figures were calculated by us from the two published blueprints; the AI-103 value uses the midpoint of Microsoft’s published range
- Published Indian and US AI engineer, AI architect and agentic specialist compensation ranges, used for role context only
360 Digital Transformation is an Anthropic Partner and a Microsoft Authorized Training Partner. Certification exams are administered independently by Anthropic and Microsoft and are not included in programme tuition. Exam details are accurate as of 20 August 2026; always confirm current format, pricing, validity and blueprint version with the issuer before booking.
Claude API
Python
Message Batches API
LangGraph
MCP
JSON Schema / Pydantic
GitHub
TypeScript
