Claude Certified Architect (CCA-F) Foundations Course 2026
Free 7-lesson course for the Claude Certified Architect (CCA-F) exam: Agent SDK, MCP, Claude Code, structured output, plus a 60-question mock exam.
Anthropic's Claude Certified Architect – Foundations (CCA-F) certification validates something the industry badly needs a signal for: the ability to make sound architectural tradeoff decisions when building production applications with Claude. Not "can you call the API" — but can you decide when a hook beats a prompt instruction, when a coordinator should re-delegate, and when batch processing is the wrong answer no matter how much it saves.
The exam tests foundational knowledge across four core technologies — Claude Code, the Claude Agent SDK, the Claude API, and the Model Context Protocol (MCP) — through 60 scenario-based questions drawn from real customer use cases: support resolution agents, multi-agent research pipelines, CI/CD code review, and structured data extraction.
This course takes you through every exam domain in order, ending with a full-length timed mock exam scored the way the real exam is scored.
What You'll Learn
The course is structured as 7 lessons mapped directly to the official exam domains, plus a practice question bank and interactive mock exam:
| Lesson | Topic | Exam Domain (Weight) |
|---|---|---|
| 1 | Agentic Loops & the Claude Agent SDK | Domain 1 — Agentic Architecture (27%) |
| 2 | Multi-Agent Orchestration & Hooks | Domain 1 — Agentic Architecture (27%) |
| 3 | Tool Design & MCP Integration | Domain 2 — Tool Design & MCP (18%) |
| 4 | Claude Code Configuration & Workflows | Domain 3 — Claude Code (20%) |
| 5 | Prompt Engineering & Structured Output | Domain 4 — Structured Output (20%) |
| 6 | Context Management & Reliability | Domain 5 — Context & Reliability (15%) |
| 7 | Exam Guide & Preparation Strategy | All domains |
| — | 60 Practice Questions with Explanations | All domains |
| — | Interactive Mock Exam (timed, scaled scoring) | All domains |
Exam at a Glance
| Detail | Value |
|---|---|
| Credential | Claude Certified Architect – Foundations |
| Questions | 60 multiple choice (1 correct, 3 distractors) |
| Time limit | 120 minutes |
| Structure | 4 scenarios drawn at random from a bank of 6 |
| Passing score | 720 on a 100–1,000 scaled scale |
| Cost | $125 USD |
| Delivery | Online proctored or test center |
| Validity | 12 months |
The distinctive feature of this exam is its scenario structure: instead of 60 unrelated questions, you get 4 production scenarios (customer support agent, Claude Code development workflow, multi-agent research system, developer productivity tooling, CI/CD integration, or data extraction), each framing a block of questions. Every question asks for the best answer among four plausible ones — the distractors are engineered to catch candidates with incomplete production experience.
Who This Course Is For
- Solution architects designing production applications on Claude
- Senior engineers building agentic systems with the Claude Agent SDK
- Platform and DevOps engineers integrating Claude Code into team workflows and CI/CD
- AI engineers who want a structured, exam-validated map of production Claude patterns
Anthropic recommends 6+ months of hands-on experience with Claude APIs, the Agent SDK, Claude Code, and MCP. This course pairs well with Anthropic Academy, the official free training platform whose 13 courses cover the same domains — use Academy for guided fundamentals and this course for exam-focused tradeoff judgment, practice questions, and a timed mock exam. If you're earlier in your journey, start with our Claude API Complete Guide and Claude Tool Use tutorial first — this course assumes you've built things and focuses on the judgment layer the exam actually tests.
Prerequisites
- Working knowledge of the Claude Messages API (requests, responses,
stop_reason) - Familiarity with tool use and JSON schemas
- Some exposure to Claude Code as a daily driver (slash commands, CLAUDE.md)
- Basic understanding of what MCP is — see Claude Model Context Protocol
How the Course Maps to the Exam
The exam weights five domains. Study time should roughly follow the weights:
- Agentic Architecture & Orchestration — 27%. The largest domain. Agentic loop control flow, coordinator–subagent patterns, context passing, hooks, workflow enforcement, task decomposition, session management. Covered in Lessons 1–2.
- Claude Code Configuration & Workflows — 20%. CLAUDE.md hierarchy, path-scoped rules, skills and slash commands, plan mode vs direct execution, CI/CD integration. Covered in Lesson 4.
- Prompt Engineering & Structured Output — 20%. Explicit criteria, few-shot examples, tool_use schemas, validation-retry loops, batch processing, multi-pass review. Covered in Lesson 5.
- Tool Design & MCP Integration — 18%. Tool descriptions, structured errors, tool distribution, MCP server configuration, built-in tool selection. Covered in Lesson 3.
- Context Management & Reliability — 15%. Long-conversation context, escalation patterns, error propagation, confidence calibration, provenance. Covered in Lesson 6.
Study Plan (3 Weeks)
Week 1 — Agentic core. Lessons 1–3. Build the hands-on exercise from each lesson; the exam rewards people who have actually watched a coordinator mis-decompose a task.
Week 2 — Tooling and output. Lessons 4–5. Configure a real project with CLAUDE.md, rules, and a skill. Build one extraction pipeline with a validation-retry loop.
Week 3 — Reliability + exam prep. Lesson 6, then the exam guide, then the practice questions. Sit the timed mock exam under real conditions (120 minutes, no notes). Score 780+ consistently before booking.
Why the CCA-F Certification Matters in 2026
Anthropic launched the Claude Certification Program in March 2026 alongside a $100 million investment in the Claude Partner Network — the first time a major AI lab has created formal credentials for its platform. That context matters for two reasons. First, partner organizations have direct commercial incentive to employ certified architects, which is what turns a certificate into interview shortlists. Second, unlike generic "AI fundamentals" certificates that test vocabulary, the CCA-F tests implementation judgment against Anthropic's actual production stack: the Agent SDK, MCP servers, and Claude Code.
The 12-month validity period is unusual — AWS gives you 3 years, TOGAF never expires — and it is a deliberate signal. The Claude platform ships breaking improvements quarterly; a certification that outlived the platform knowledge it validates would be worthless. Budget for annual recertification the way you budget for dependency upgrades.
If you already hold an architecture credential like the AWS Solutions Architect Associate or TOGAF, you will find the exam style familiar but the failure modes different: cloud exams punish you for not knowing service limits, while the CCA-F punishes you for over-engineering — reaching for a classifier or a routing layer when a richer tool description would do.
How Scoring Actually Works
Your result is a scaled score from 100 to 1,000, with 720 required to pass. Scaled scoring means Anthropic statistically equates different exam forms, so the raw number of correct answers needed shifts slightly with form difficulty — on a typical form, 720 corresponds to roughly 42 of 60 questions. Three practical consequences:
- There is no partial credit and no penalty for guessing. The platform forces an answer before you can advance, so a disciplined elimination process on hard questions directly buys points.
- Domain performance is invisible in the pass/fail result, so diagnose weaknesses with practice tools before exam day — our mock exam reports per-domain accuracy for exactly this reason.
- Time pressure is mild but real: 2 minutes per question sounds generous until a scenario stem runs 150 words. Practice reading scenario context once, then answering its whole question block without re-reading.
Start the Course
Begin with Lesson 1: Agentic Loops & the Claude Agent SDK — the foundation every other domain builds on.
Frequently Asked Questions
What does the Claude Certified Architect – Foundations certification cover?
Five weighted domains: Agentic Architecture & Orchestration (27%), Claude Code Configuration & Workflows (20%), Prompt Engineering & Structured Output (20%), Tool Design & MCP Integration (18%), and Context Management & Reliability (15%). Every question is scenario-based — you are dropped into a production context like a support agent or a CI pipeline and asked to choose the best architectural response among four plausible options. That is why this course pairs every lesson with hands-on exercises rather than flashcard facts.
How long does it take to prepare for the CCA-F exam?
Anthropic recommends 6+ months of hands-on experience with the Claude API, Agent SDK, Claude Code, and MCP before attempting the exam. On top of that foundation, most candidates need 15–20 focused hours — which is what this course's 3-week plan schedules. If you are new to Claude entirely, budget 30–40 hours and work through Anthropic Academy's free courses first, using this course afterward for exam-specific tradeoff judgment.
Is this Claude certification course free?
Yes — all 7 lessons, the 60-question practice bank, and the timed interactive mock exam are free, with no signup. The official exam itself costs $125, paid to Anthropic when you book through their certification portal. Anthropic Academy's 13 official preparation courses are also free, so the only unavoidable cost of getting certified is the exam fee.
Do I need to complete Anthropic Academy before this course?
No. Academy teaches the underlying skills; this course assumes them and drills the exam's judgment layer. If you can build an agent with tool calling today, start here and dip back into Academy for any domain where the mock exam shows a gap.