CCA Foundations Exam Domains at a Glance
This page is the fast-reference version of the CCA Foundations exam structure: what's on it, how much each part is worth, and which prep resource maps to which domain.
Use it to plan study time before diving into any single domain in depth.
How to Use This Cheatsheet
- Scan the domain weight table first and allocate study hours proportionally, not evenly.
- Use the format facts table to understand exam-day constraints before you start practice questions.
- Follow the "map to this section" column to jump to the deeper article for whichever domain you're studying.
- Revisit this page in the final week before the exam as a coverage checklist, not a first-read.
Exam Format Facts
| Fact | Detail |
|---|---|
| Certification name | Claude Certified Architect (CCA) - Foundations |
| Launch date | 2026-03-12 |
| Question count | 60 questions |
| Delivery | Proctored |
| Book policy | Closed-book |
| Assistance policy | No AI assistance permitted during the exam |
| Tests | Ability to design and ship production-grade Claude applications at enterprise scale |
| Free prep material | Anthropic Academy, Skilljar-hosted, launched 2026-03-02 |
Domain Weights
| Domain | Weight | Core focus | Map to this section |
|---|---|---|---|
| Agentic Architecture & Orchestration | 27% | Multi-agent design, orchestration patterns, agent-to-agent handoff | Studying Agentic Architecture and Orchestration |
| Claude Code Configuration & Workflows | 20% | Hooks, subagents, custom commands, repeatable team workflows | Tool Design, MCP Integration, and Context Management Checklist |
| Prompt Engineering & Structured Output | 20% | Reliable, parseable output at scale; prompt design discipline | What the Exam Actually Tests |
| Tool Design & MCP Integration | 18% | Tool schemas, MCP server integration, safe tool surface area | Tool Design, MCP Integration, and Context Management Checklist |
| Context Management & Reliability | 15% | Token budgets, context window discipline, long-run failure modes | Tool Design, MCP Integration, and Context Management Checklist |
Total: 100%. Agentic Architecture & Orchestration is the single largest domain; the next two (Claude Code Configuration & Workflows, Prompt Engineering & Structured Output) are tied at 20% each; Tool Design & MCP Integration and Context Management & Reliability round out the remaining 33% combined.
Domain-by-Domain Quick Facts
Agentic Architecture & Orchestration (27%)
| Aspect | Detail |
|---|---|
| Weight rank | 1st (largest domain) |
| Core question shape | When to use single-agent vs. multi-agent, supervisor-worker vs. peer topologies, handoff design |
| Common failure tested | Agent loops that never terminate, poor task decomposition, unclear ownership between agents |
| Related Academy course | Building with the Claude API |
Claude Code Configuration & Workflows (20%)
| Aspect | Detail |
|---|---|
| Weight rank | Tied 2nd |
| Core question shape | Configuring hooks, subagents, custom commands, and CI-integrated workflows for a team |
| Common failure tested | One-off manual workflows that don't scale to a team, misconfigured permission/hook boundaries |
| Related Academy course | Claude Code in Action |
Prompt Engineering & Structured Output (20%)
| Aspect | Detail |
|---|---|
| Weight rank | Tied 2nd |
| Core question shape | Designing prompts and output schemas that stay reliable across many calls, not just one |
| Common failure tested | Output that parses in testing but breaks under edge-case input at scale |
| Related Academy course | Building with the Claude API |
Tool Design & MCP Integration (18%)
| Aspect | Detail |
|---|---|
| Weight rank | 4th |
| Core question shape | Tool schema design, MCP server boundaries, least-privilege tool surface area |
| Common failure tested | Tools that return unbounded output, overly broad tool permissions, ambiguous tool descriptions |
| Related Academy course | Introduction to MCP |
Context Management & Reliability (15%)
| Aspect | Detail |
|---|---|
| Weight rank | 5th (smallest domain by weight) |
| Core question shape | Token budget planning, context window discipline over long agent runs |
| Common failure tested | Silent context overflow, stale context carried across turns, unbounded conversation history |
| Related Academy course | Building with the Claude API |
Applying This Cheatsheet
- Weeks 1-2: study Agentic Architecture & Orchestration and Claude Code Configuration & Workflows together, since they're the two largest weight blocks (27% + 20% = 47% of the exam).
- Weeks 3-4: study Prompt Engineering & Structured Output and Tool Design & MCP Integration; these two domains frequently overlap in real question scenarios.
- Final week: review Context Management & Reliability specifically, since its failure modes (context overflow, token mismanagement) show up as root causes inside questions nominally about other domains.
FAQs
Which single domain should get the most study time?
Agentic Architecture & Orchestration, at 27% it is the largest of the five domains and its concepts recur inside questions framed around other domains too.
Do the five domain weights add up to 100%?
Yes: 27 + 20 + 20 + 18 + 15 = 100%.
Are Claude Code Configuration & Workflows and Prompt Engineering & Structured Output really tied in weight?
Yes, both are weighted at 20% each, making them jointly the second-largest share of the exam behind Agentic Architecture & Orchestration.
Is Context Management & Reliability safe to study last since it's the smallest domain?
Not entirely safe to skip until last: at 15% it's the smallest by direct weight, but its failure modes (context overflow, token budget mismanagement) frequently appear as the root cause inside questions nominally scored under other domains.
What exactly does "no AI assistance" mean on exam day?
It means no use of Claude, Claude Code, any other AI assistant, or AI-powered search during the proctored exam session itself.
Are the Anthropic Academy courses required before sitting the exam?
They aren't stated as a strict prerequisite, but the free Skilljar-hosted courses (launched 2026-03-02) are built to map onto exam domain themes and are a reasonable primary study path.
How many questions come from each domain?
The exam doesn't publish a fixed per-domain question count; the percentages describe the proportional weight of each domain's contribution to the score, which is a reasonable proxy for how many questions are drawn from it across 60 total.
Is this exam open-book?
No, it's closed-book, in addition to prohibiting AI assistance, which makes it stricter than most day-to-day Claude workflows.
Which domain covers MCP specifically?
Tool Design & MCP Integration, weighted at 18%, covers Model Context Protocol integration alongside general tool schema design.
Which domain covers multi-agent handoffs?
Agentic Architecture & Orchestration, the 27% domain, covers multi-agent design and orchestration patterns including handoffs between agents.
Is there a domain specifically about Claude Code hooks and subagents?
Yes, Claude Code Configuration & Workflows (20%) covers hooks, subagents, custom commands, and team-level workflow configuration.
Should I study domains in weight order or in some other order?
Weight order is a reasonable default, but studying overlapping domains together (for example, Tool Design & MCP Integration alongside Context Management & Reliability) can be more efficient since real questions often blend them.
Related
- What the Claude Certified Architect Exam Actually Tests - the conceptual mental model behind these domain weights and the exam's format.
- Studying Agentic Architecture and Orchestration: the CCA Exam's Largest Domain - a full treatment of the 27% domain summarized here.
- Tool Design, MCP Integration, and Context Management: CCA Domain Checklist - enumerated study points for three of the five domains listed above.
- CCA Foundations Best Practices - practical guidance for applying all five domains to real builds.
Stack versions: Written against the Claude model lineup current as of ~June 2026 - Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5 (the default), and Claude Haiku 4.5 - and the official
anthropicPython SDK (latest 0.x release). Model names, pricing, exam format, and SDK versions move quickly - verify current specifics at platform.claude.com/docs and the official CCA exam guide before relying on them.