Claude Certified Architect – Foundations (CCA-F) Study Guide

32 Articles · 8 Tasks

Foundations

F1 Messages API (5)

F1.1

The Messages API Is Stateless — Every Request Stands Alone

Messages API Request Structure

arrow_forward

F1.2

Two Roles, No Exceptions

Message Roles

arrow_forward

F1.3

Four Content Block Types, Each With a Direction

Content Block Types

arrow_forward

F1.4

Reading the Response: stop_reason Is Your Control Signal

Response Structure

arrow_forward

F1.5

Tokens: The Currency of Every API Call

Tokens and Context Window

arrow_forward

F2 Tool Use (5)

F2.1

The Tool Use Dance: Six Steps, Three API Calls

Tool Use Complete Flow

arrow_forward

F2.2

Defining Tools: Name, Description, and Input Schema

Tool Definition Format

arrow_forward

F2.3

The tool_use Block: What the Model Sends When It Wants a Tool Called

tool_use Content Block

arrow_forward

F2.4

Sending Tool Results Back: The tool_result Block

tool_result Content Block

arrow_forward

F2.5

Virtual Tools: Hijacking tool_use for Guaranteed JSON

tool_use as Structured Output

arrow_forward

F3 MCP (5)

F3.1

MCP: One Protocol to Connect Them All

What is MCP

arrow_forward

F3.2

Tools, Resources, and Prompts: MCP's Three Primitives

MCP Three Primitives

arrow_forward

F3.3

The camelCase Trap: MCP Tool Definitions vs Claude API

MCP Tool Definition Format

arrow_forward

F3.4

Don't Disguise Errors as Empty Results

MCP CallToolResult

arrow_forward

F3.5

stdio for Local, HTTP for Remote: Choosing MCP Transports

MCP Server Transport Methods

arrow_forward

F4 Claude Code (4)

F4.1

Claude Code: A CLI Tool, Not an IDE

What is Claude Code

arrow_forward

F4.2

Grep Searches Content, Glob Searches Paths

Built-in Tools

arrow_forward

F4.3

Three Levels of CLAUDE.md: User, Project, Directory

Configuration Hierarchy

arrow_forward

F4.4

Four Permission Modes: From Full Control to Full Automation

Permission Modes

arrow_forward

F5 Agent SDK (5)

F5.1

Agent SDK: Claude Code's Agentic Powers, Programmable

What is Agent SDK

arrow_forward

F5.2

query() vs ClaudeSDKClient: Stateless vs Stateful

Core Interfaces

arrow_forward

F5.3

ClaudeAgentOptions: Configuring What an Agent Can Do

ClaudeAgentOptions

arrow_forward

F5.4

When Statefulness Actually Matters

query vs ClaudeSDKClient

arrow_forward

F5.5

Sub-Agents: Coordinator Delegates, Specialists Execute

Sub-Agents in Agent SDK

arrow_forward

F6 JSON Schema (2)

F6.1

Seven Types: string, number, integer, boolean, array, object, null

JSON Schema Core Types

arrow_forward

F6.2

required, nullable, and enum: Schema Features That Shape Output

Schema Features

arrow_forward

F7 Prompt Basics (3)

F7.1

System Prompt: Persistent Frame, Not Per-Turn Input

System Prompt vs User Message

arrow_forward

F7.2

Few-Shot: Show, Don't Just Tell

Few-shot vs Zero-shot

arrow_forward

F7.3

Temperature: Determinism Is Not Accuracy

Temperature

arrow_forward

F8 Agentic Concepts (3)

F8.1

Agents Do Things; Chatbots Talk About Things

Agent vs Chatbot

arrow_forward

F8.2

The Agentic Loop: Reason → Act → Observe → Repeat

Agentic Loop

arrow_forward

F8.3

Five Orchestration Patterns: From Simple Chains to Dynamic Coordination

Orchestration Patterns

arrow_forward