Claude Certified Architect PrepGet certified in 6 months · start now
Sign inSign up free
Concepts

27 technical primitives, made scannable.

Concepts are the 27 technical primitives the CCA-F (Claude Certified Architect - Foundations) exam tests on. Each page lowers friction with TLDR stats, diagrams, evidence pips, exam traps, and a sticky study rail. Grouped by exam domain. New here? See how this fits, or jump to the Claude Certified Architect certification guide.

27 concepts
Not sure where to start? Take the 10-min diagnostic
Agentic loops - painterly hero featuring Loop mascot at the four-stage runtime.

Featured concept

Agentic loops: think, act, observe, decide. The mascot becomes the concept anchor.

Template LOpen page ↗
Domain D1 · Agentic Architectures

D1 · 27% of the exam.

7 concepts in this domain.

D1.1

Agentic Loops

An agentic loop turns one API response into a controlled, multi-turn system. The harness reads stop_reason, di…

D1.2

Subagents

Subagents are specialized, isolated agents spawned by a coordinator to handle domain-specific tasks while pres…

D1.3

Stop Reason

stop_reason is the authoritative struct field that says why Claude stopped, either end_turn (finished) or tool…

D1.4

Session State & Persistence

Session state is the running message list (and any external persistence) that gives an agentic loop continuity…

D1.5

Human-in-the-Loop Escalation

Escalation is the deterministic handoff path when policy thresholds, low confidence, or repeated failure condi…

D1.6

Subagent State Handoff

Subagents hand off state three ways: structured messages (default - typed JSON envelopes through the orchestra…

D1.7

Agent Instruction Files (CLAUDE.md / AGENTS.md)

CLAUDE.md is Claude-specific; AGENTS.md is the emerging cross-tool convention (Codex, Cursor, Aider). Portabil…

Domain D2 · Tool Design + Integration

D2 · 18% of the exam.

5 concepts in this domain.

D2.1

Tool Calling

Tool calling is how Claude decides to invoke external functions and pass structured arguments. Tools are routi…

D2.2

Tool Choice

tool_choice is the parameter that controls whether Claude can decide to use a tool ("auto"), must call any too…

D2.3

Model Context Protocol

MCP is a communication standard that lets Claude access pre-built tools, resources, and prompts from specializ…

D2.4

SDK Hooks (Pre/PostToolUse)

Hooks are deterministic code that runs before or after tool calls. They enforce policy that prompt-only patter…

D2.5

Tool Evaluation & Testing

Evaluation is how you measure tool-call correctness against ground-truth datasets. A full deep-dive guide is c…

Domain D3 · Agent Operations

D3 · 20% of the exam.

4 concepts in this domain.

D3.1

CLAUDE.md Hierarchy

CLAUDE.md is persistent project memory loaded automatically by Claude Code. A three-level hierarchy (user → pr…

D3.2

Plan Mode

Plan Mode forces Claude Code to draft and approve a multi-step plan before executing. Use it for non-trivial c…

D3.3

Skills

Skills are reusable, on-demand task workflows defined in markdown with YAML frontmatter. Unlike CLAUDE.md (alw…

D3.4

Checkpoints & Session Management

Checkpoints save and restore conversation state. A full deep-dive guide is coming soon. {🟡 Claude Code checkp…

Domain D4 · Prompt Engineering

D4 · 20% of the exam.

9 concepts in this domain.

D4.1

System Prompts & Instructions

System prompts establish role, constraints, format, and tool guidance. Anatomy: role, capability boundaries, s…

D4.2

Prompt Caching

Prompt caching reduces cost (~90%) on repeated context like long system prompts and tool definitions. Cache br…

D4.3

Batch API

Message Batches API: 50% discount for async, non-time-sensitive workloads. A full deep-dive guide is coming so…

D4.4

Structured Outputs

Structured outputs guarantee Claude returns JSON matching a schema instead of natural language. Force a tool w…

D4.5

Vision & Multimodal

Vision lets Claude process images alongside text. A full deep-dive guide is coming soon. {🟢 messages content …

D4.6

Streaming

Streaming returns tokens as they're generated for low-latency UX. A full deep-dive guide is coming soon. {🟢 m…

D4.7

Attention Engineering

Attention engineering is the discipline of placing critical context where the model attends most strongly, hig…

D4.8

4D Framework

Delegation, Description, Discernment, Diligence, Anthropic's 4D framework for agent prompt design. A full deep…

D4.9

Prompt Engineering Techniques

Prompt engineering is a craft of seven techniques that turn a brittle one-shot prompt into a production-grade …

Domain D5 · Context + Reliability

D5 · 15% of the exam.

2 concepts in this domain.

D5.1

Context Window Management

Context window management is how you keep long conversations within limits without dropping critical facts. Pa…

D5.2

Case Facts Block

A case-facts block is an immutable set of transactional data (customer ID, order details, refund amount, polic…

FAQ

6 questions about the concept catalog

Every Q is phrased as a real Google search query. Answers cite the same evidence-tagged sources used elsewhere on the site.

What are 'concepts' on this site?
27 single-page deep-dives on individual Claude technical primitives - the building blocks the CCA-F exam tests. Each concept covers: definition, mental model, code example, common mistakes, links to related scenarios + knowledge pages. Concepts are the smallest atom in the curriculum - scenarios show how 3-7 concepts combine into a production architecture.
How are concepts organized?
Grouped by the 5 CCA-F domains: D1 Agentic Architectures (agentic loops, subagents, stop_reason, session state, escalation, hooks), D2 Tool Design + Integration (tool calling, tool choice, MCP), D3 Agent Operations (evaluation, plan mode, skills, checkpoints), D4 Prompt Engineering (system prompts, prompt caching, batch API, structured outputs), D5 Context + Reliability (vision/multimodal, streaming, attention engineering, 4D framework, context window, case-facts block, claude.md hierarchy, prompt engineering techniques).
Which concept has the highest exam weight?
Agentic loops (in D1, the 27%-weighted domain) carries the highest individual question density - roughly 6-8 of the 60 exam questions assume working knowledge of the loop pattern, stop_reason branching, and how subagents fit. Tool calling (D2) and prompt engineering techniques (D4) are the next-highest individual density.
Where should I start studying concepts?
Take the 10-Q diagnostic first to identify your weakest domain. The diagnostic result shows the recommended concept-to-start for each domain. Most candidates should begin with agentic loops (the foundational concept the rest of D1 builds on) regardless of weak spots, then drill the weakest domain second.
Do all 27 concepts have the same depth?
No - depth scales with exam weight. High-weight concepts (agentic loops, tool calling, MCP) have full deep-dives with multiple code examples, mental-model diagrams, and 5+ anti-pattern callouts. Lower-weight concepts (e.g., streaming, batch API) have shorter pages focused on the 1-2 things the exam actually tests.
How are concepts linked to scenarios and knowledge pages?
Three-way cross-link. Each concept page has a 'where this is used' rail listing the 2-4 scenarios that depend on it. Each concept also links to the matching Anthropic Academy summary in the Knowledge pillar. Reverse links: each scenario page lists its prerequisite concepts as 'primitives' at the top.