How to Pass the Anthropic Claude Certified Architect (CCA-F) in 15 Days: The Complete Roadmap

Originally published on CertLand.

On March 12, 2026, Anthropic did something no AI-native company had done before: they launched a professional certification that tests whether you can actually_build_with their platform, not just talk about it. The*Claude Certified Architect — Foundations (CCA-F)is the first vendor certification from an AI lab — not a cloud hyperscaler, not a training company — and it has already become one of the most discussed credentials in the AI engineering community. If you are a developer, solutions architect, or technical lead who builds applications with Claude, this certification validates the exact skills employers are hiring for right now. This guide gives you a complete 15-day study plan to pass it on your first attempt. **In this guide:*

1. What Is the CCA-F Certification
2. Exam Format and Logistics
3. The 5 Exam Domains with Weights
4. 15-Day Study Plan (Day by Day)
5. Top 5 Exam Traps Candidates Fall Into
6. Free Study Resources
7. Practice with 450 Questions on CertLand
8. Final Tips for Exam Day

What Is the CCA-F Certification

The Claude Certified Architect — Foundations (CCA-F) is Anthropic's first professional certification. It validates that you can design and implement production-grade applications using Claude's core technologies: the*Claude Agent SDK,Claude Code, theClaude API(Messages API, extended thinking, batch processing), and theModel Context Protocol (MCP)*. Unlike generic AI certifications that test theoretical machine learning knowledge, the CCA-F is grounded in real-world scenarios: building customer support agents, designing multi-agent research pipelines, integrating Claude Code into CI/CD workflows, extracting structured data from messy documents, and orchestrating tool calls across backend systems.

Anthropic designed the exam around six realistic production scenarios, of which four are randomly selected for each exam sitting. This means every candidate sees a different combination of scenarios, making question-bank memorization useless and practical experience essential. The certification targets solution architects with at least six months of hands-on experience building with Claude APIs, the Agent SDK, Claude Code, and MCP.

Exam Format and Logistics Detail Value Exam name Claude Certified Architect — Foundations (CCA-F) Launch date March 12, 2026 Number of questions 60 (all single-choice) Time limit 120 minutes Passing score 720 / 1000 (scaled) Cost $99 per attempt (free for Anthropic Partner Network members) Delivery Online proctored Penalty for guessing No penalty — answer every question

The scaled scoring model (100 to 1000) means different exam forms are equated for difficulty. A 720 does not mean you need exactly 72% correct — it means you met the minimum standard set by Anthropic's subject matter experts across the scaled difficulty curve. In practice, aim for 75%+ on your practice exams to have a comfortable margin.

The 5 Exam Domains with Weights

The CCA-F covers five content domains. The weights tell you exactly where to invest your study time. Domain 1 alone accounts for more than a quarter of the exam — if you are weak on agentic architecture, you cannot pass. Domain Topic Weight ~Questions 1 Agentic Architecture & Orchestration 27% ~16 2 Tool Design & MCP Integration 18% ~11 3 Claude Code Configuration & Workflows 20% ~12 4 Prompt Engineering & Structured Output 20% ~12 5 Context Management & Reliability 15% ~9

Domain 1: Agentic Architecture & Orchestrationcovers the Claude Agent SDK, orchestrator-subagent patterns, multi-agent delegation, lifecycle hooks, tool handoff protocols, agent failure recovery, and human-in-the-loop escalation. This is the highest-weighted domain because agentic AI is the fastest-growing application pattern in 2026, and Anthropic wants architects who can design these systems safely.

Domain 2: Tool Design & MCP Integrationtests your ability to design Model Context Protocol tool and resource interfaces, define tool schemas, handle tool execution results, configure MCP servers, and integrate backend systems (databases, APIs, file systems) through the MCP layer. Expect questions on when to use MCP resources versus tools, and how to structure tool descriptions so Claude selects the right tool reliably.

Domain 3: Claude Code Configuration & Workflowscovers CLAUDE.md files (project-level, user-level, and enterprise-level), custom slash commands (Agent Skills), plan mode versus direct execution, MCP server integration in Claude Code, CI/CD pipeline integration for automated code review and test generation, and permission management.

Domain 4: Prompt Engineering & Structured Outputtests system prompt design, few-shot prompting, chain-of-thought reasoning, JSON schema enforcement, XML-tagged outputs, extraction patterns for unstructured documents, extended thinking configuration, and batch API usage for high-throughput processing.

Domain 5: Context Management & Reliabilitycovers context window management across long documents and multi-turn conversations, token budgeting, retrieval-augmented generation (RAG) integration, self-evaluation patterns, error handling strategies, retry logic, and monitoring and observability for production Claude deployments.

15-Day Study Plan (Day by Day)

This plan assumes you have some familiarity with Claude (you have used the API or Claude Code at least a few times). If you are starting from zero, add three to five days to the front for basic orientation. Each day targets roughly 2 to 3 hours of focused study. The plan is designed to work with CertLand's Coach feature, which links each study session to targeted practice questions from our 450-question CCA-F bank. Day Focus Activities 1 Domain 1a:Agent SDK Foundations Study the Claude Agent SDK architecture: agent lifecycle, system prompts for agents, tool binding, and the orchestrator pattern. Read Anthropic Academy's "Building Agentic Systems" module. Complete 25 Domain 1 practice questions on CertLand. 2 Domain 1b:Multi-Agent & Escalation Deep dive into multi-agent orchestration: subagent delegation, context handoff between agents, human-in-the-loop workflows, failure recovery, and lifecycle hooks. Complete remaining Domain 1 practice questions. 3 Domain 2a:MCP Fundamentals Study the Model Context Protocol specification: tools vs. resources, transport layers (stdio, SSE), mcp.json configuration, and tool schema design. Read the official MCP documentation. Complete 20 Domain 2 practice questions. 4 Domain 2b:Tool Design Patterns Practice designing MCP tool schemas, writing clear tool descriptions, handling tool errors, and chaining multiple tool calls. Build a simple MCP server locally if possible. Complete remaining Domain 2 questions. 5 Domain 3a:Claude Code Setup Study CLAUDE.md hierarchy (project, user, enterprise), permission scopes, the .claude/ directory structure, slash commands, and Agent Skills. Install Claude Code and experiment with a real project. Complete 20 Domain 3 practice questions. 6 Domain 3b:CI/CD & Plan Mode Study plan mode vs. direct execution, CI/CD integration (automated code review, test generation, PR feedback), MCP servers in Claude Code, and headless mode. Complete remaining Domain 3 questions. 7 Domain 4a:Prompt Engineering Core Study system prompt design, role prompting, few-shot examples, chain-of-thought reasoning, and prompt chaining. Review Anthropic's prompt engineering documentation. Complete 20 Domain 4 practice questions. 8 Domain 4b:Structured Output & Batch API Study JSON schema enforcement, XML-tagged outputs, data extraction patterns, extended thinking (budget tokens, reasoning visibility), and the batch API for high-throughput use cases. Complete remaining Domain 4 questions. 9 Domain 5a:Context & Token Management Study context window limits, token budgeting strategies, conversation summarization, the /compact command, and managing context across multi-agent handoffs. Complete 15 Domain 5 practice questions. 10 Domain 5b:RAG & Reliability Study RAG integration patterns, self-evaluation and self-correction loops, retry logic, error handling, monitoring and observability for production deployments. Complete remaining Domain 5 questions. 11 Full Simulation #1 Take a full 60-question timed practice exam on CertLand (120 minutes). Do not look up answers during the simulation. Score yourself and identify your weakest domains. 12 Full Simulation #2 Take a second full simulation with different questions. CertLand's 450-question bank ensures minimal overlap. Compare results against Simulation #1 to measure improvement. 13 Targeted Review — Weak Domains Use CertLand's domain filter to drill into your weakest areas. Re-read explanations for every question you got wrong. Focus on understanding_why_the correct answer is correct, not just memorizing it. 14 Targeted Practice — Hard Questions Use CertLand's difficulty filter to focus on hard-difficulty questions across all domains. These scenario-based questions most closely mirror the real exam's complexity. Aim for 75%+ accuracy. 15 Final Review + Exam Day Light review only: skim your notes on each domain, re-read the official exam guide, review your most-missed question categories. Rest well. Take the exam with confidence. 💡 Pro Tip: Do not skip the two full simulations on Days 11 and 12. The CCA-F is a 120-minute exam with scenario-based questions that require careful reading. Time management is a real factor — you have exactly 2 minutes per question on average. Practicing under timed conditions is the single most effective thing you can do to avoid running out of time on exam day.

Top 5 Exam Traps Candidates Fall Into

Based on early candidate feedback and our analysis of the exam guide, these are the five most common traps that catch unprepared candidates. Knowing these exists gives you a significant edge.

Trap 1: Confusing MCP tools with MCP resources.The exam tests whether you know when to use a tool (an action that modifies state, like processing a refund) versus a resource (read-only data retrieval, like fetching a customer profile). Many candidates treat them as interchangeable — the exam does not.

Trap 2: Choosing the most capable solution instead of the most appropriate one.When a question asks how to handle a simple classification task, the answer is not "build a multi-agent orchestrator with extended thinking." The exam rewards architectural restraint — using the simplest pattern that meets the requirements.

Trap 3: Ignoring token economics in context management questions.Questions about long-document processing and multi-turn conversations test whether you understand the cost and latency implications of context window usage, not just whether you can fit everything into the window.

Trap 4: Defaulting to Claude Opus for every scenario.The exam includes questions where Claude Haiku or Claude Sonnet is the correct choice because the scenario prioritizes speed, cost, or high-throughput over maximum reasoning capability. Always read the scenario constraints carefully.

Trap 5: Overlooking human-in-the-loop requirements.Domain 1 heavily tests escalation patterns. If a question describes a scenario where the agent is making high-stakes decisions (financial transactions, account deletions), the correct answer almost always includes a human approval gate — not an autonomous action.

Free Study Resources

Anthropic has made it possible to prepare for the CCA-F without spending money on courses. Here are the essential free resources: Resource What It Covers URL Anthropic Academy Official free courses covering prompt engineering, tool use, agentic workflows, and the Agent SDK. Directly aligned with CCA-F domains. anthropic.skilljar.com Official Exam Guide (PDF) Domain breakdown, task statements, sample questions, and all six exam scenarios described in detail. Available on Anthropic's certification page Claude Documentation API reference, prompt engineering guide, tool use guide, extended thinking, and batch API documentation. docs.anthropic.com MCP Specification The full Model Context Protocol spec, including tool schemas, resource definitions, and transport configuration. modelcontextprotocol.io Claude Code Documentation CLAUDE.md configuration, slash commands, Agent Skills, plan mode, and CI/CD integration. docs.anthropic.com/en/docs/claude-code 💡 Pro Tip: Start with the Anthropic Academy courses at anthropic.skilljar.com. They are free, self-paced, and built by the same team that wrote the exam. The "Prompt Engineering" and "Real World Agents" courses map almost directly to Domains 1 and 4. If you only have time for two resources, use the Academy and the official exam guide PDF.

Practice with 450 Questions on CertLand

Reading documentation builds knowledge. Practice questions build exam readiness. They are not the same thing. The CCA-F uses scenario-based questions with plausible distractors that test judgment, not recall — and the only way to develop that judgment under time pressure is to practice with realistic questions and study the explanations. CertLand CCA-F Practice Exam — 450 Questions

CertLand offers the most comprehensive CCA-F practice exam available —450 scenario-based questionscovering all five domains, with detailed explanations for every answer. Each explanation tells you why the correct answer is right_and_why every distractor is wrong, with exam tips to help you remember key patterns.

• 450 questionsacross all 5 domains, weighted to match the real exam
• Difficulty filtering— practice easy, medium, or hard questions independently
• Domain filtering— drill into your weakest areas
• Coach-guided study plans— let CertLand's Coach feature guide you through the 15-day plan above with targeted question sets for each session
• Timed simulation mode— practice under real exam conditions (60 questions, 120 minutes)
• Full explanations— every question explains all four options plus an exam tip

Start practicing now atCertLand.net

The 15-day study plan above is designed to work hand-in-hand with CertLand's question bank. On each study day, after reviewing the material for a domain, immediately practice with the corresponding questions on CertLand. This active recall pattern — study, then test, then review explanations — is backed by decades of cognitive science research and is far more effective than passive reading alone.

Final Tips for Exam Day

The CCA-F rewards practical judgment over memorization. Here is how to maximize your score on exam day:

Read the scenario first, then the question.The CCA-F uses scenario-based questions where four questions share a single production scenario. Read the scenario carefully before attempting any of the associated questions — the constraints described in the scenario are essential context for choosing the correct answer.

Eliminate two distractors, then choose.Most CCA-F questions have two clearly wrong answers and two plausible ones. Quickly eliminate the obvious distractors, then focus your reasoning on the remaining two options. This approach is faster and more accurate than evaluating all four options equally.

Watch for "most appropriate" language.The exam often asks for the "most appropriate" or "best" approach, not the only correct one. Multiple answers may technically work, but one is more aligned with Anthropic's recommended patterns, more cost-effective, or simpler. Choose the option that best fits the specific constraints in the scenario.

Do not leave questions blank.There is no penalty for guessing. If you are stuck, eliminate what you can and select your best guess. A 25% chance is better than 0%.

Budget your time.120 minutes for 60 questions gives you 2 minutes per question. If a question is taking more than 3 minutes, flag it and move on. Come back to flagged questions after completing the rest of the exam.

The CCA-F is a credential that signals you can build real applications with Claude — not just talk about AI in the abstract. With 15 days of focused preparation, the right study resources, and consistent practice on CertLand, you can pass it on your first attempt. Start your 15-day countdown today.