All 54 CCAR‑F lessons
The complete Claude Certified Architect – Foundations study guide, one page per lesson. Free, sourced to Anthropic’s own documentation, with diagrams and comprehension checks.
Domain 1: Agentic Architecture & Orchestration
27% of exam · 12 lessons · ~54 min
1
The agentic loop and stop_reasonEvery agent you build on the Messages API is the same four-step loop, and one response field drives it.
5 min · diagram · table · 2 checks
2
pause_turn and server-side toolsServer-side tools run their own sampling loop. When it hits its limit you get a pause, not an error.
3 min · 2 checks
3
Parallel tool use and result pairingOne assistant message can request several tools. How you return the results decides whether it ever does so again.
3 min · 2 checks
4
Workflow patterns: chaining, routing, parallelisationThree of the five canonical patterns. The exam tests which one fits a described workload.
6 min · diagram · table · 2 checks
5
Orchestrator-workers and evaluator-optimizerThe two patterns for work you cannot fully specify up front.
4 min · diagram · 2 checks
6
Coordinator and subagent architectureThe single most-tested idea in the heaviest domain: subagents do not inherit context.
6 min · diagram · 2 checks
7
Subagent prompting and tool scopingHow you brief a subagent, and what you let it touch.
4 min · 2 checks
8
Session state, resume and crash recoveryWhat to do when a long pipeline dies halfway.
4 min · 2 checks
9
Agent or workflow? Surfaces, tiers and effortWhen an agent is justified at all, which surface to build it on, and how to tune spend.
5 min · table · 2 checks
10
Error classification and propagationNot all failures are the same kind, and the kind decides who handles them.
5 min · table · 2 checks
11
Fallback, retry and graceful degradationWhat the system does when the happy path is unavailable.
4 min · 2 checks
12
State: in-context versus external memoryWhere an agent keeps what it knows, and why the choice matters.
5 min · 2 checks
Domain 2: Claude Code Configuration & Workflows
20% of exam · 11 lessons · ~53 min
1
Settings files and precedenceFive scopes, one order, and one important exception.
5 min · diagram · table · 2 checks
2
Permissions: enforcement vs guidanceThe distinction the exam returns to again and again.
4 min · table · 2 checks
3
The CLAUDE.md hierarchy and load orderWhere memory files live, in what order they load, and why nested files behave differently.
6 min · diagram · table · 2 checks
4
Path-scoped rules with .claude/rules/How to keep instructions out of context until they are relevant.
4 min · 2 checks
5
Imports, AGENTS.md and exclusionsComposing memory files across repos and tools.
4 min · 2 checks
6
Hooks: events, exit codes and decisionsThe enforcement layer. Exit codes are near-guaranteed exam content.
6 min · diagram · table · 2 checks
7
MCP config and CI/CD workflowsSharing servers with a team, and running Claude Code without a human.
4 min · 2 checks
8
Skills and custom slash commandsThe two used to be separate features. They have merged, and the exam-relevant details are the frontmatter and the precedence.
6 min · 2 checks
9
Claude Code architecture and the execution modelWhat actually runs where, and why that determines what you can enforce.
5 min · 2 checks
10
Excluding files and controlling project scopeHow to keep Claude away from files it should not read, and what actually enforces it.
4 min · 2 checks
11
The Agent SDK and headless automationRunning the same harness programmatically, and in CI where nobody can answer a prompt.
5 min · 2 checks
Domain 3: Prompt Engineering & Structured Output
20% of exam · 12 lessons · ~53 min
1
Structured outputs: two distinct featuresConstraining the response body and constraining tool parameters are separate knobs.
5 min · 2 checks
2
JSON Schema support and limitsWhat the schema engine will and will not enforce.
4 min · table · 2 checks
3
Syntax errors vs semantic errorsThe highest-value idea in this domain: schemas guarantee shape, never truth.
5 min · diagram · 2 checks
4
Prefill removal and its replacementsA migration detail that shows up as a scenario.
3 min · table · 2 checks
5
Prompt construction for productionWhere each kind of content belongs, and why placement is a caching decision too.
4 min · 2 checks
6
Message Batches: economics and limitsNumbers worth memorising, and the one workload Batches must never serve.
5 min · 2 checks
7
Batch recovery and accuracy measurementHandling partial failure, and the statistics trap before you automate.
4 min · 2 checks
8
Structuring complex prompts with XML tagsWhen a prompt carries several kinds of content, delimiters stop them bleeding into each other.
4 min · 2 checks
9
Generation parameters and what replaced themtemperature, top_p and top_k are gone on current models. Knowing what to use instead is the exam-relevant part.
5 min · 2 checks
10
Multi-turn design and context accumulationConversations grow, and how you let them grow determines both cost and coherence.
5 min · 2 checks
11
Evaluating prompts: datasets, grading and regressionHow you know a prompt change was an improvement rather than a vibe.
5 min · 2 checks
12
Chain-of-thought and extended thinkingGetting the model to reason before answering, and how that is configured now.
4 min · 2 checks
Domain 4: Tool Design & MCP Integration
18% of exam · 10 lessons · ~46 min
1
MCP architecture: host, client, serverThe participant model and the two layers, which the exam states precisely.
4 min · 2 checks
2
MCP primitives and transportsThree server primitives, and knowing which one a scenario actually calls for.
5 min · table · 2 checks
3
Tool descriptions are the primary leverWhen the model picks the wrong tool, the description is almost always the fix.
4 min · 2 checks
4
Designing tool inputs and outputsThe cluster of design rules the exam tests most heavily.
6 min · 2 checks
5
Tool errors and trust boundariesWho fixes which failure, and what you are allowed to believe.
4 min · 2 checks
6
Scaling the tool surfaceWhat to do when there are too many tools to put in context.
4 min · 2 checks
7
Building an MCP serverWhat you actually implement, and the decisions that shape the surface you expose.
5 min · 2 checks
8
MCP clients and multi-server routingWhat the host does when several servers are connected at once.
4 min · 2 checks
9
Idempotency, partial success and safe retriesThe property that decides whether retrying a tool is safe at all.
5 min · 2 checks
10
MCP security and production hardeningA server is an interface a model drives. Scope it accordingly.
5 min · 2 checks
Domain 5: Context Management & Reliability
15% of exam · 9 lessons · ~43 min
1
Prompt caching mechanicsOne invariant explains every caching question on the exam.
5 min · diagram · 2 checks
2
Cache economics and verificationThe numbers, and how to prove caching is actually working.
4 min · table · 2 checks
3
Invalidation hierarchy and silent invalidatorsNot every change invalidates everything — and the ones that do are predictable.
5 min · table · 2 checks
4
Context editing, compaction and memoryThree mechanisms, three different jobs. Expect a "which one" question.
4 min · diagram · table · 2 checks
5
Long-context effectsTwo distinct failure modes with two distinct signatures.
4 min · 2 checks
6
Provenance, conflict and escalationReliability practices that keep multi-agent output trustworthy.
5 min · 2 checks
7
RAG architecture: chunking, embedding, retrievalThe pipeline that puts the right context in front of the model instead of all of it.
6 min · 2 checks
8
Semantic, keyword and hybrid retrievalTwo retrieval methods with complementary blind spots, and why production uses both.
5 min · 2 checks
9
Production RAG: multi-index, citation and groundingWhat separates a demo pipeline from one you can put in front of users.
5 min · 2 checks