Anthropic Certification

Claude Certified Architect
Foundations — Study Guide

Everything you need to pass the Claude Certified Architect Foundations exam. Free, complete, scenario-based.

5
Exam Domains
6
Scenarios
MCQ
Format
6mo+
Experience Needed

What is it

About the Certification

The Claude Certified Architect – Foundations certification validates that you can make informed decisions about tradeoffs when implementing real-world solutions with Claude. It tests foundational knowledge across Claude Code, the Claude Agent SDK, the Claude API, and Model Context Protocol (MCP).

Questions are scenario-based, drawn from actual customer use cases including agentic systems for customer support, multi-agent research pipelines, CI/CD integrations, developer productivity tools, and structured data extraction.

Target candidate: 6+ months of practical experience building with Claude APIs, Agent SDK, Claude Code, and MCP.

Exam Blueprint

5 Exam Domains

All scored content falls into these five domains.

25%
Domain 1
Agentic Architecture & Orchestration
  • Agentic loop lifecycle (tool_use vs end_turn)
  • Multi-agent coordination patterns
  • Task decomposition strategies
  • Human-in-the-loop workflows
  • Subagent spawning via Task tool
20%
Domain 2
Tool Design & MCP Integration
  • Model Context Protocol (MCP) architecture
  • Tool schema design and validation
  • MCP server configuration
  • Tool result handling and error patterns
  • allowedTools configuration
20%
Domain 3
Claude Code Configuration & Workflows
  • CI/CD pipeline integration
  • Claude Code hooks and permissions
  • Automated code review workflows
  • Test generation patterns
  • CLAUDE.md configuration
20%
Domain 4
Prompt Engineering & Structured Output
  • System prompt design for agents
  • Structured output with JSON schemas
  • Chain-of-thought prompting
  • Few-shot examples in production
  • Anti-patterns to avoid
15%
Domain 5
Context Management & Reliability
  • Context window management
  • Memory and state patterns
  • Error handling and retries
  • Self-evaluation patterns
  • Escalation design

Exam Format

6 Exam Scenarios (4 appear on your exam)

Each exam presents 4 randomly selected scenarios from these 6. Questions are multiple choice with one correct answer.

1
Customer Support Resolution Agent
Agentic ArchitectureTool Design & MCPContext Management
2
Multi-Agent Research Pipeline
Agentic ArchitecturePrompt EngineeringContext Management
3
Structured Data Extraction System
Prompt EngineeringTool Design & MCPReliability
4
Developer Productivity with Claude
Tool Design & MCPClaude CodeAgentic Architecture
5
Claude Code for CI/CD
Claude CodeAgentic ArchitectureReliability
6
Enterprise Agent Orchestration
Agentic ArchitectureContext ManagementTool Design & MCP

Exam Prep

Key Things to Know

🔄
Agentic Loop
Know the difference between stop_reason "tool_use" (continue) and "end_turn" (stop). This appears frequently.
🔧
Task Tool
allowedTools must include "Task" for a coordinator to spawn subagents. Subagent context must be explicitly provided.
🔒
Hooks over Prompts
When business rules require guaranteed compliance, use programmatic hooks — not prompt instructions alone.
🧠
Context Pollution
Know strategies for keeping context clean across sessions. Unbounded memory growth is a common anti-pattern.
📡
MCP Architecture
Understand MCP server setup, tool schema design, and how tool results flow back into conversation context.
📋
CLAUDE.md
Know how CLAUDE.md configures Claude Code behavior, permissions, and project-specific instructions.

Built by a Practitioner

This guide is maintained by Sameer Goel — AI builder and solutions architect with extensive experience building autonomous agent systems. The content reflects real patterns from production agent deployments.

Visit sameer-goel.com →