How to Pass the Claude Certified Architect Exam: Complete Study Guide

The Claude Certified Architect — Foundations (CCA-F) is Anthropic's first professional certification. It tests whether you can make sound architectural decisions when building with Claude — not just recite documentation, but choose correctly between valid approaches in ambiguous scenarios. This guide covers everything you need to prepare effectively.

Exam Structure at a Glance

• Format: Proctored, multiple-choice (single-select and multi-select)
• Duration: 60 minutes
• Questions: 27–28 per attempt (weighted sample across domains)
• Scoring: 1000-point weighted scale
• Pass mark: 720 / 1000
• Eligibility: Corporate email from an Anthropic Partner Network organisation
• Cost: $99 (often free for qualifying partner tiers)

The 5 Domains and Weightings

The exam is not evenly distributed. Knowing which domains carry the most weight is the first study strategy decision:

• Domain 1 — Agentic Architecture & Orchestration: 27%
  The largest domain. Covers the agentic loop, single vs multi-agent patterns, coordinator/specialist architecture, tool design, and the Agent SDK.
• Domain 3 — Claude Code Configuration & Workflows: 20%
  CLAUDE.md configuration, hooks, plugins, slash commands, and the MCP integration within Claude Code.
• Domain 4 — Prompt Engineering & Structured Output: 20%
  System prompts, XML structuring, few-shot examples, structured outputs, output validation, and prompt injection defense.
• Domain 2 — Tool Design & MCP Integration: 18%
  Tool schema design, the isError pattern, MCP server/client/transport architecture, tool safety, and injection defense.
• Domain 5 — Context Management & Reliability: 15%
  Context window management, summarisation strategies, RAG integration, prompt caching, and reliability patterns.

What the Exam Actually Tests

The CCA-F is explicitly designed to be harder than a documentation quiz. Anthropic built it around architectural judgment traps — scenarios where multiple answers are technically valid, but only one is contextually correct.

Common trap patterns you will face:

• System prompt logic vs. programmatic hard gates (soft enforcement vs. guaranteed enforcement)
• Adding few-shot examples when the real fix is a better tool description
• Self-review by the same session vs. independent instance review
• Over-engineering with classifiers before trying simpler deterministic approaches
• Message Batches API inside blocking workflows (batches have 24-hour processing windows)
• Checking response text content instead of reading stop_reason directly
• Too many tools per agent (>4–5 causes silent reliability degradation)
• Progressive summarisation in scenarios that require verbatim precision

Study Strategy: 35–40 Hours

Candidates with prior Claude experience typically need 35–40 hours of focused preparation. If you're starting from scratch, plan for 5–10 weeks. Here's how to allocate that time:

1. Weeks 1–2: Domain 1 deep dive (Agentic Architecture)
   Build a simple agent with tools. Read the Agent SDK documentation. Understand the agentic loop lifecycle, tool call flow, and multi-agent communication patterns. Do at least one practice exam focused on Domain 1.
2. Weeks 3–4: Domains 3 and 4 (Claude Code + Prompt Engineering)
   Install Claude Code and use it daily. Configure a real CLAUDE.md for a project. Build one plugin. Practice system prompt design across different scenarios. Study structured output usage.
3. Week 5: Domains 2 and 5 (MCP + Context Management)
   Build or connect to an MCP server. Study transport types and the three capability types (tools, resources, prompts). Review context management strategies and the summarisation trap.
4. Final week: Full simulation exams
   Take all practice exams. Focus on time pressure — 60 minutes for 27+ questions means ~2 minutes per question. Review every question you got wrong and understand why the correct answer is correct.

Official Study Resources

• Anthropic Academy — Free official courses including Claude 101, Building with Claude API, MCP integration, and Claude Code.
• Anthropic Documentation — The canonical reference. Pay particular attention to the tool use guide, agent patterns section, and the MCP documentation.
• MCP Specification — Essential for Domain 2. Read the architecture overview and the server implementation guide.

Practice Exams vs Passive Study

Reading documentation alone will not close the gap between knowledge and architectural judgment. The CCA-F's trap-based question format requires you to practice choosing between plausible options under time pressure. Candidates who supplement their reading with practice exams consistently outperform those who read more but test less.

The most effective revision loop: take a timed practice exam → review all incorrect answers → re-read the relevant documentation section → immediately take another practice exam on the same domain. Repeat until you're consistently scoring above 750 on practice attempts.