Registrations close Tuesday, 18 August 2026Lectures open Thursday, 20 August 2026
Vizuara · Claude Partner Network

The Vizuara preparation course for the CCAF exam.

Recorded lectures, assignments, a 55,000-word book, and a question bank covering all five domains of the Claude Certified Architect – Foundations exam. Work through it at your own pace and finish certified by Vizuara, a member of the Claude Partner Network.

Recorded lectures · lifetime access Assignments per domain CCAF book · 117 figures Question bank Vizuara certification Limited time · one intake
Enrol now · ₹25,000 See the curriculum ↓

Registrations close on 18 August 2026. Lectures become available on 20 August 2026. This is preparation training by Vizuara AI Labs; the official CCAF credential is issued by Anthropic.

Your certificate comes from a Claude Partner Network member
Registrations close 18 August. Lectures open 20 August.
This is a limited time course. Anthropic is building out more certification tracks, and we will follow them with further courses in the CCAF track. This intake, at this price, closes on 18 August.
–
Days
–
Hours
–
Minutes
–
Seconds
01 · Why this certificate carries weight

Vizuara is a member of the Claude Partner Network.

Anthropic runs a partner network for organisations that build and teach on Claude. Vizuara is a member of it. That membership is what sits behind the certificate you receive at the end of this course.

You are certified by Vizuara, not by Anthropic

On completion you receive a Vizuara certification. It is not the Anthropic CCAF credential, and we will not pretend otherwise. What it says is that you completed a full preparation programme run by a Claude Partner Network member.

The material tracks the official exam guide

Partner access means we teach from the current exam guide rather than from guesswork. All five domains, their published weightings, the scenario format, and the preparation exercises are covered as Anthropic describes them.

You finish ready to sit the real exam

The point of the course is the Anthropic exam: 60 scenario questions, 120 minutes, 720 out of 1,000 to pass. Everything here is built so that you can book that exam and walk in prepared.

Vizuara joins the Claude Partner Network
Anthropic · Claude Partner Network member
02 · What you get

Four things, and they are the four that actually move your score.

There are no live sessions in this course. Everything is recorded and written, so you can move at whatever pace your week allows and come back to any part of it later.

01 · Recorded lectures

Recorded lectures across all five domains

Every domain taught scenario first, because that is how the exam asks. Watch at your own speed, rewind the parts that matter, keep access for good.

  • Five modules, one per exam domain
  • Lecture time weighted to match the exam weightings
  • Available from 20 August 2026
  • Lifetime access, no expiry on the recordings
02 · Assignments

Assignments you build, not just read

Each domain closes with an assignment drawn from the preparation exercises in the exam guide. People who have built the thing pass; people who only read about it tend not to.

  • Structured extraction pipeline with a retry loop
  • Multi-agent research coordinator with subagents
  • MCP tool descriptions tested against ambiguous requests
  • A CLAUDE.md and rules setup for a real repository
03 · The CCAF book

A 55,000-word book written for this exam

One chapter per domain, with 117 hand-drawn figures. It is the reference you keep open beside the lectures and the thing you reread in the last week before the exam.

  • Five chapters, roughly 11,000 words each
  • 117 figures explaining the mechanisms visually
  • Reads in the browser on phone or laptop
  • Yours for as long as the course runs
04 · Question bank

A question bank in the exam's own format

Scenario questions with four plausible answers and one correct call, the way the real paper is written. Every question carries an explanation of why the other three fail.

  • Questions grouped by domain for targeted drilling
  • Full-length timed sets in the 60-question format
  • Explanations for the correct answer and each distractor
  • Domain-level scoring so you can see where you are weak
On the certification: completing the course earns you a certification issued by Vizuara AI Labs. It is not the Anthropic CCAF credential, which is earned only by passing Anthropic's proctored exam. What Vizuara's certification represents is completion of a preparation programme run by a Claude Partner Network member.
03 · The exam

What the Claude Certified Architect exam actually tests.

The Foundations exam checks whether you can make architectural tradeoffs when building production systems with Claude Code, the Claude Agent SDK, the Claude API, and MCP. Every question is a scenario: a real production context, four plausible answers, one correct call.

60
Questions
120
Minutes
720
Pass score /1000
$125
Exam fee
5
Content domains
12mo
Validity
1

Agentic Architecture & Orchestration

Agentic loops, coordinator and subagent patterns, hooks, task decomposition, session state and forking.

27% of exam
2

Tool Design & MCP Integration

Tool descriptions that route correctly, structured errors, tool distribution, .mcp.json scoping, built-in tools.

18% of exam
3

Claude Code Configuration & Workflows

CLAUDE.md hierarchy, rules and skills, slash commands, plan mode against direct execution, CI/CD integration.

20% of exam
4

Prompt Engineering & Structured Output

Explicit criteria, few-shot design, JSON schemas via tool_use, validation-retry loops, batch strategies.

20% of exam
5

Context Management & Reliability

Long-context strategies, escalation patterns, error propagation, human review, provenance.

15% of exam
04 · Curriculum

Five recorded modules, one per exam domain.

Each module pairs recorded lectures with the matching book chapter, an assignment, and a domain set in the question bank. Work through them in order, or go straight to the domain you are weakest in.

MODULE 01Domain 4 · 20% of the exam

Prompt Engineering & Structured Output

We start here because it is the fastest domain to secure points in, and because it teaches you how the exam's distractors are built.

Covered in the lectures

  • Scenario anatomy: how questions are constructed and how distractors are engineered
  • The agentic loop, stop_reason of tool_use against end_turn, and tool results in context
  • Prompts with explicit criteria, and why categorical rules beat vague instructions
  • Few-shot design for ambiguous cases and reducing false positives
  • Structured output through tool_use and JSON schemas, with tool_choice modes
  • Nullable fields against hallucinated values, and resilient enum plus other-detail schemas
  • Validation-retry loops and the Message Batches API

Assignment

Structured data extraction pipeline

Build an extraction tool with a resilient schema, a validation-retry loop that feeds errors back, and a routing decision between batch and realtime.

Book ch. 4Question bank set 4
MODULE 02Domain 1 · 27% of the exam

Agentic Architecture & Orchestration

The heaviest domain on the paper. Taught through failure modes, because the exam asks what broke and what fixes it.

Covered in the lectures

  • Hub-and-spoke coordination: decomposition, delegation, aggregation of results
  • Subagent context isolation, and why findings have to be passed explicitly in prompts
  • Parallel subagents through multiple Task calls in a single response
  • Hooks for deterministic compliance, against prompt-based guidance
  • Prompt chaining against dynamic decomposition, and narrow-decomposition failures
  • Session management with resume and fork_session, and stale-context tradeoffs
  • Escalation and handoff protocols built on structured summaries

Assignment

Multi-agent research pipeline

Build a coordinator with search and analysis subagents: parallel spawning, structured claim and source output, error propagation with partial results, and a refinement loop.

Book ch. 1Question bank set 1
MODULE 03Domain 2 · 18% of the exam

Tool Design & MCP Integration

Tools are how Claude touches your systems. Descriptions, contracts, errors, and MCP configuration.

Covered in the lectures

  • Tool descriptions as the routing mechanism: inputs, edge cases, boundaries, when to use
  • Splitting overloaded tools, and renaming to remove functional overlap
  • Structured MCP errors with isError, errorCategory and isRetryable
  • Tool distribution: why loading 18 tools onto one agent fails and scoped access works
  • MCP scoping across project .mcp.json and user config, with env-var expansion for secrets
  • MCP resources for content catalogues, and community servers against custom builds
  • Built-in tools: Read, Write, Edit, Bash, Grep and Glob, and when each is correct

Assignment

Design and break MCP tools

Write tool descriptions that route correctly under ambiguous requests, add structured error responses, then adversarially test selection reliability.

Book ch. 2Question bank set 2
MODULE 04Domain 3 · 20% of the exam

Claude Code Configuration & Workflows

How a real team configures Claude Code, and how the exam tests whether you know which lever to pull.

Covered in the lectures

  • The CLAUDE.md hierarchy across user, project and directory, and imports
  • Rules with glob scoping, and when a rule beats an instruction in the prompt
  • Skills and slash commands, including context forking, allowed-tools and argument hints
  • Plan mode against direct execution, and the Explore subagent for verbose discovery
  • CI and CD: headless runs, JSON output, schema-constrained output, review instances
  • Choosing between a hook, a rule, a skill and a subagent for the same requirement

Assignment

Configure a repository end to end

Write the CLAUDE.md hierarchy, scoped rules, one skill and one slash command for a real codebase, then wire a headless review step into CI.

Book ch. 3Question bank set 3
MODULE 05Domain 5 · 15% of the exam

Context Management, Reliability & Exam Strategy

The last domain, then the paper itself: timing, elimination, and the full-length timed set.

Covered in the lectures

  • Long-context strategies, lost-in-the-middle effects, case-facts blocks and scratchpads
  • Compaction and what it costs you
  • Escalation calibration, error propagation, confidence scoring and human review routing
  • Provenance: keeping track of which claim came from where
  • Exam strategy: time budgeting across 60 questions, and eliminating distractors
  • The proportionate first response heuristic, which decides more questions than it should

Assignment

Full-length timed set

Sit a 60-question set in the exam format and time limit, scored on the 100 to 1,000 scale, then work through a domain-level gap analysis.

Book ch. 5Question bank · full set
05 · The CCAF book

Fifty-five thousand words, written for this exam and nothing else.

Most exam preparation sends you to scattered documentation. This course ships a single book: one chapter per domain, written to be read straight through, with figures that show the mechanism rather than decorate the page.

54,973
Words
117
Figures
5
Chapters, one per domain
∞
Rereads before the exam

Each chapter runs to roughly eleven thousand words and follows the same shape: introduce the idea, show it working, then say what it means for a question on the paper. It reads in the browser, on a phone or a laptop, and it is the thing to reread in the last week before you sit the exam.

Chapter 2 · Tool Design & MCP Integration

Why a tool description is a routing decision

When Claude has more than one tool available, the description is the only thing it can use to choose between them. A description that reads well to a human, but does not say what the tool refuses to do, will route the wrong request into it. The fix is rarely a better model and almost always a boundary written into the description.

request route search_docs search_code
Figure 2.4 · The description is the router

A page from the book, shortened

06 · Question bank

Questions written the way the exam writes them.

The difficulty in this exam is not recall. It is that all four answers look reasonable and three of them are wrong for a specific reason. The question bank drills exactly that, and every explanation says why each distractor fails.

Drill one domain at a time

Questions are grouped by domain, so after a lecture you can go straight to the set that matches it and find out whether it landed.

Then sit the full length

Full sets run in the exam's format: 60 questions, 120 minutes, scored on the 100 to 1,000 scale with 720 to pass.

Every distractor explained

Getting a question right by luck teaches nothing. Each answer carries a note on why it is correct or where exactly it goes wrong.

Domain 1 · Sample
A research coordinator spawns three subagents. Each returns a correct summary, but the coordinator's final report contradicts itself on a key figure. What is the most likely cause?
AThe subagents ran in parallel rather than in sequence
BSubagent findings were not passed into the coordinator's prompt, so it reconciled from memory
CThe context window was exceeded and older messages were dropped
DThe coordinator needed a larger model
Why B. Subagent contexts are isolated. The coordinator only sees what is returned to it explicitly, so a report that reconciles figures it never received is a context-passing bug, not a capacity or a model problem.
07 · Your certification

Certified by Vizuara, a Claude Partner Network member.

Finish the five modules and their assignments and you receive a Vizuara certification, issued under our membership of the Claude Partner Network.

To be plain about what it is and is not: this certification comes from Vizuara. Anthropic issues the official Claude Certified Architect credential, and the only way to hold that one is to pass Anthropic's proctored exam. This course is built to get you to that exam prepared.

The certification carries your name, the domains covered, and the date of completion, and it is verifiable through Vizuara.

VIZUARA · CLAUDE PARTNER NETWORK Certificate of Completion CCAF Exam Preparation Course All five domains · assignments completed ISSUED BY VIZUARA AI LABS
A limited time course

This is the first course in the CCAF track, and it runs once at this price.

Anthropic is building out its certification programme, and more tracks are coming. We intend to follow them with further courses as they arrive, which means this intake is the first of several rather than the only one. What it will not be is repeated on these terms. Registrations close on 18 August, and the next intake opens later in the year at a higher price.

08 · Pricing

One price. Everything included.

CLOSES 18 AUG
Recorded preparation course
₹25,000 / seat

Registrations close 18 August 2026. Lectures open 20 August 2026.

  • Recorded lectures across all five exam domains, with lifetime access
  • Assignments for every domain, built from the official preparation exercises
  • The CCAF book: 54,973 words, 117 figures, five chapters
  • Question bank with domain sets and full-length timed sets
  • Explanations for every correct answer and every distractor
  • Vizuara certification on completion
Enrol now · ₹25,000

The certification is issued by Vizuara AI Labs and is not the official Anthropic CCAF credential. The $125 Anthropic exam fee is separate and paid to Anthropic.

Enterprise

Teams and partner organisations

For engineering teams putting several people through the course together, write to us and we will scope it separately.

  • Bulk seats for your engineering org
  • Progress and readiness reporting per engineer
  • Guidance on Claude Partner Network membership
  • GST invoicing and purchase order support
Contact us →
Dr. Raj Dandekar
09 · Your instructor

Dr. Raj Dandekar

PhD, MIT · Founder, Vizuara AI Labs

Raj builds AI-powered developer tools and agentic systems at Vizuara, and has taught live cohorts on modern software development, LLM inference engineering, and voice agents to thousands of engineers.

This course comes out of months of building production Claude systems: multi-agent pipelines, MCP integrations, and Claude Code workflows in CI. The exam does not test what the features are. It tests when and why you would reach for each one, and that is what the lectures spend their time on.

10 · FAQ

Questions, answered straight.

Are there any live sessions?
No. This course is entirely recorded. There are no live lectures and no fixed class times. Lectures become available on 20 August 2026 and you work through them at your own pace, with lifetime access.
When do registrations close, and when do I get access?
Registrations close on Tuesday, 18 August 2026. Lectures, the book, the assignments and the question bank become available on Thursday, 20 August 2026. If you enrol before the deadline you get access on the 20th.
What does "Vizuara is a Claude Partner Network member" mean for me?
Anthropic runs a partner network for organisations that build and teach on Claude, and Vizuara is a member of it. In practice it means the course is taught from the current exam guide, and the certification you receive at the end comes from a partner organisation rather than from an unaffiliated training provider.
Will I be certified by Anthropic at the end of this?
No. You will be certified by Vizuara. The official Claude Certified Architect credential is issued by Anthropic and is earned only by passing Anthropic's proctored exam, which costs $125 and is paid to Anthropic separately. This course prepares you for that exam and gives you a Vizuara certification for completing it.
What exactly is included?
Four things: recorded lectures covering all five exam domains, an assignment for each domain, the CCAF book at 54,973 words with 117 figures, and a question bank with domain sets and full-length timed sets. Completion earns the Vizuara certification.
What are the prerequisites?
Comfort with Python or TypeScript and some exposure to LLM APIs. Anthropic recommends around six months of hands-on Claude experience before sitting the exam. The course compresses a good part of that through the assignments, but you should be a working developer. Module 1 starts from the platform fundamentals.
How long does it take to get through?
Most people work through it over three to four weeks alongside a job, spending a few hours a week per module. There is no deadline on your side once you have access, so you can take longer if you need to.
Can anyone sit the official Anthropic exam?
Eligibility and availability for the Anthropic exam are set by Anthropic and have been changing as the programme opens up. Check anthropic.com for the current position before you book. Preparation takes weeks and eligibility can change quickly, so the useful move is to be ready.
What is the refund policy?
Because the course is delivered as recorded material and a book, refunds are available up to the point access is granted on 20 August 2026. After access is granted we do not offer refunds, though you keep lifetime access to everything in the course.
Registrations close 18 August 2026

Be ready for the exam by the time you are eligible to sit it.

Recorded lectures, assignments, the book, and the question bank. Certified by Vizuara, a member of the Claude Partner Network. Lectures open on 20 August.

Enrol now · ₹25,000