Claude Certified Architect Exam: Format, Domains, Cost, and How to Prepare

I've had three cups of Java, and I still don't understand object-oriented programming! — Me, probably in the '90s.

Anthropic just announced its first official technical certification for Claude Code called Claude Code Architect, Foundations. The exam, only open to partner companies at this time, is focused on solution architects who use Claude Code.

I've been reading the study guide published by Anthropic to get an idea of what it's all about, and to help separate the facts from the misinformation. This article should fill in the blanks and help you discover what is covers, how to take it and, most importantly, how you can prepare now so you're ready when it's released to the general public.

What Is the Claude Certified Architect, Foundations Certification?

As I mentioned above, the Claude Certified Architect, Foundations exam (or CCA), is Anthropic's first official certification targeted to solution architects. The exam is not a memory-recall-type of exam. It's a proctored, scenario-based test that will judge your knowledge of architectural decisions for Claude-based production systems.

Anthropic has also confirmed there will be many other certifications in the future to meet you where you are in terms of knowledge and skill, such as credentials for sellers, developers, and advanced architects. While a specific date hasn't been given, they are expected to arrive later this year.

Who Is It For?

The Claude Certified Architect, Foundations exam is targeted toward solution architects building applications in production with Claude. Anthropic has classified this as a "301" level exam, which means it assumes prior knowledge of Claude, its API, and its tools. If you're building workflows in Claude Code, designing agentic systems, or making architectural decisions, this exam is for you.

Even if you don't think you're among the targeted audience, studying this material will certainly help with your knowledge of Claude Code. The exam domains cover the most important skills in the Claude ecosystem.

Exam Format and Details

Here are the exam details:

60 multiple-choice questions, scenario-based
120 minute time limit
The exam is proctored. No Claude, documentation, or external tools will be allowed
Each question has one correct answer out of four options
There are six possible scenarios. Four of which will be randomly selected for your exam
Passing score: 720 out of 1,000
Results are delivered within two working days
If you fail, a retake requires a mandatory waiting period and a new exam purchase

Again, this exam is about applying your knowledge, not testing your memorization.

How Much Does It Cost?

The exam is only available to the first 5,000 partner company employees who get early access at no cost. After that, the exam fee is $99.

How Do You Access It?

The Claude Certified Exam is only available thorugh the Claude Partner Network. The network is free to join for any organization bringing Claude to market, such as consultant organizations, AI companies, developer shops, etc.

You can register through the Skilljar platform at anthropic.skilljar.com.

What Does the Exam Cover?

The certification exam is based around five domains, each with different weights. The scoring mechanism has not been released by Anthropic to-date, although you must score 720/1000 to pass.

Domain 1: Agentic Architecture and Orchestration (27%)

This is the largest weighted category of all five domains. It covers designing multi-agent systems, managing task decomposition, implementing agentic loops, hub-and-spoke orchestration patterns, hooks, session management, and the Agent SDK. If you're developing Claude-based systems around autonomous actions, this domain is one to know.

Domain 2: Tool Design and MCP Integration (18%)

This domain, fourth smallest in terms of weight, covers creating Model Context Protocol servers, managing tool boundaries, how to structure tool descriptions, handle errors, and work with Claude's built-in tools.

Domain 3: Claude Code Configuration and Workflows (20%)

A middle-of-the-road domain in terms of weight. It covers the CLAUDE.md hierarchy, custom slash commands and skills, plan mode, iterative refinement strategires, CI/CD integration, and batch processing. If you've been reading the articles on this site, you're already becoming familiar with the concepts in this domain.

Domain 4: Prompt Engineering and Structured Output (20%)

Another middle-of-the-road domain in terms of weight. This section covers JSON schema design, the tool_use patter, validation-retry loops, few-shot prompting, field-level confidence scoring, and human review workflows. This domain is going to test whether you can build reliable, structured outputs from a non-deterministic model.

Domain 5: Context Management and Reliability (15%)

And the lightest weighted domain covers preserving long-context information, managing handoff patterns between agents, confidence calibration, and ensuring system reliability at scale.

The Six Scenarios

The exam uses scenario-based questions from real-world contexts. There are six possible scenarios, and your exam will randomly select four of the following:

A customer support agent using the Agent SDK with escalation logic
A Claude Code team workflow with CLAUDE.md configuration and plan mode
A multi-agent research system with parallel task orchestration
A developer tools project using Claude Agent SDK with MCP servers
A CI/CD pipeline integration using the -p flag and batch processing
A structured data extraction pipeline with validation-retry loops

You won't know which four you get until the exam starts. That means you need to be comfortable with all six.

How to Prepare

Anthropic offers free preparation through the Anthropic Academy on Skilljar. There are 13 self-paced courses covering the full Claude ecosystem, including:

Claude 101 (foundational course)
Building Applications with the Claude API (8.1 hours, the flagship course)
Model Context Protocol (MCP) Advanced
AI Fluency: Frameworks and Foundations
Claude Code on Coursera (part of Vanderbilt's Generative AI specialization)

All of these courses are free and open to everyone.

Study These First

Claude Code Hooks (Domain 1 — highest weight at 27%) Agents and Subagents (Domain 1) Programmatic Prerequisites (Domain 1) CLAUDE.md Hierarchy (Domain 3) Path-Specific Rules (Domain 3) Context Window (Domain 5)

Beyond the official courses, hands-on experience matters most. Build something with Claude Code. Design an agentic workflow. Create a CLAUDE.md hierarchy for a real project. The exam tests application, not memorization.

Whether or not you plan to take the exam, understanding what's covered in the five domains will help you to become better at using and building with Claude Code. These are the most important topics in the ecosystem today.