Certifications9 min readBy Rohit Mote

The 30-Day CCA-F Study Plan: A Day-by-Day Schedule to Pass the Claude Certified Architect Exam

A day-by-day, 30-day study plan for the Claude Certified Architect (CCA-F) exam — built for working professionals studying 45-60 minutes a day.

The 30-Day CCA-F Study Plan: A Day-by-Day Schedule to Pass the Claude Certified Architect Exam

"Study for the CCA-F over four weeks" is easy advice to give and hard advice to follow. Most working professionals don't have four undefined weeks — they have 45 minutes before dinner, or a lunch break, or the first half hour after the kids are asleep. Vague timelines fail because they don't tell you what to open your laptop and do today.

This is a day-by-day plan instead. 30 days, roughly 45-60 minutes each, mapped directly to the five domains Anthropic tests on the Claude Certified Architect Foundations (CCA-F) exam. Skip a day, and you'll know exactly where you left off — no re-planning required.

Before Day 1: Set Up Your Tracking

Two things, five minutes total:

  • Get the official Exam Guide PDF from Anthropic's certification page — it lists the five domains and their weights, which this plan is built around.
  • Create a simple tracker — a spreadsheet or even a notes doc with 30 rows. Each day you'll log: topic covered, confidence (1-5), and one thing you didn't understand. This log becomes your Week 4 weak-spot list.
  • Week 1: Foundation (Days 1-7)

    The goal this week isn't mastery — it's building an accurate mental map of what you don't know yet, so the rest of the plan targets real gaps instead of comfortable topics.

    DayFocusTask
    1OrientationRead the full Exam Guide PDF. Note the five domain weights.
    2Anthropic AcademyComplete Module 1 (free, Skilljar)
    3Anthropic AcademyComplete Module 2
    4Anthropic AcademyComplete Module 3
    5Anthropic AcademyComplete Module 4
    6Hands-onBuild one small Claude Code workflow end-to-end — a CLAUDE.md, one custom skill, one hook
    7CheckpointTake the official practice exam cold, no studying beforehand. Don't aim to pass — this is your baseline score.
    Why take the practice exam this early? A cold baseline score tells you your real starting point, not your assumed one. Most candidates overestimate their readiness on Domain 2 (Claude Code configuration) and underestimate Domain 4 (MCP) — you want that surprise now, with 23 days left, not on exam day.

    Week 2: Domain 1 — Agentic Architecture & Orchestration (Days 8-14)

    This is the heaviest domain on the exam at 27% of your score, so it gets its own full week. It covers orchestrator-subagent patterns, failure handling, and the tradeoffs between parallel and sequential agent pipelines.

    DayFocusTask
    8Orchestration patternsRead Anthropic's docs on multi-agent architecture. Note when orchestrator-worker beats a single long-running agent.
    9DelegationStudy how context gets passed between agents without blowing the token budget
    10Failure handlingLearn retry logic, timeout patterns, and graceful degradation for tool call failures
    11Parallel vs. sequentialWork through 3 scenarios: when does parallelizing agents help vs. hurt?
    12Human-in-the-loopStudy checkpoint and interruption patterns — where should a human approve before an agent proceeds?
    13BuildExtend your Day 6 project into a two-agent pipeline: one orchestrator, one subagent with a narrow task
    14Self-testWrite 10 scenario questions of your own for Domain 1 and answer them. Forcing yourself to write the question is a stronger test than answering someone else's.

    Week 3: Domains 2, 3, and 4 (Days 15-24)

    Three domains, ten days — this week is denser. Domains 2 and 3 are tied at 20% each; Domain 4 is 18%. Together they're just as heavy as Domain 1, so don't let the week feel like a "lighter" one.

    DayFocusTask
    15Domain 2: CLAUDE.mdStudy the file hierarchy — root, subdirectory, project-level — and how inheritance resolves conflicts
    16Domain 2: Skills & hooksWrite one custom skill with correct frontmatter; study hook types (pre-tool, post-tool, on-error)
    17Domain 2: Permissions & CI/CDStudy permission modes and the --headless flag for pipeline integration
    18Domain 3: Prompt structurePractice role/task/context framing; study when few-shot examples help vs. hurt
    19Domain 3: Structured outputGenerate JSON from Claude with and without schema constraints — deliberately try to break it, and note the failure modes
    20Domain 3: ReliabilityStudy grounding techniques and prompt chaining vs. single long prompts
    21Domain 4: MCP fundamentalsStudy MCP tool definition anatomy — name, description, input schema
    22Domain 4: Build a serverBuild one MCP server from scratch using the TypeScript or Python SDK
    23Domain 4: Security boundariesStudy local vs. remote MCP servers and the principle of least privilege for tool authorization
    24CheckpointTake a mixed 30-question quiz covering Domains 2-4 only. Log every wrong answer in your tracker.

    Week 4: Domain 5, Weak Spots, and Simulation (Days 25-30)

    The last stretch is where your tracker from Weeks 1-3 pays off. You already know exactly where you're weak — this week targets those gaps directly instead of re-reviewing everything evenly.

    DayFocusTask
    25Domain 5: Context managementStudy token budget strategies and prompt caching — what gets cached, cache TTL, and cost impact
    26Domain 5: EvalsStudy lightweight evaluation frameworks for Claude outputs, and A/B testing prompts in production
    27Weak-spot reviewOpen your tracker. Re-study your two lowest-confidence topics from the last three weeks.
    28Weak-spot reviewRe-study your next two lowest-confidence topics
    29Full simulationTake a full 60-question practice exam under real conditions: 120 minutes, no interruptions, no notes
    30Final reviewReview every wrong answer from Day 29 until you understand why the correct answer is correct — not just what it is
    Target before scheduling the real exam: 850+/1000 on your Day 29 simulation. The passing score is 720, but scenario exams have more variance than trivia exams — a comfortable buffer protects you from an off day.

    Six Scenarios to Know Cold

    The CCA-F draws its questions from six recurring production scenarios, four of which appear on any given exam. Make sure your Week 2-3 hands-on practice touches at least three of these directly:

  • Customer Support Resolution Agent (ticket triage + resolution)
  • Code Generation with Claude Code (autonomous PR review/generation)
  • Multi-Agent Research System (orchestrator + subagents)
  • Developer Productivity with Claude (IDE integration, CLAUDE.md setup)
  • Claude Code for CI/CD (headless Claude in a build pipeline)
  • Structured Data Extraction (documents into validated JSON schemas)
  • Frequently Asked Questions

    Is 45-60 minutes a day actually enough?

    Yes, for candidates who already build with Claude in some form — API, Claude Code, or agent pipelines. If you're starting from zero hands-on experience, budget closer to 90 minutes on the "Build" and "Hands-on" days (6, 13, 22) since those require actually writing and running code, not just reading.

    Can I compress this into two weeks?

    You can, but double up on domain-review days rather than cutting the two full practice exams (Day 7 and Day 29) — those are the highest-signal checkpoints in the plan and the first thing candidates regret skipping when they retake the exam.

    What if I score below 720 on the Day 29 simulation?

    Don't schedule the real exam yet. Add a Week 5: spend 3-4 days on your two lowest-scoring domains specifically, then run a fresh full simulation before booking. Scheduling before you're consistently clearing 800 just converts a study problem into a $200 mistake.

    Do I need to memorize the token limits, pricing, or model names?

    No. The CCA-F is explicitly scenario-based — it tests architectural judgment, not trivia. Time spent memorizing numbers that change with product updates is better spent on the "why" behind each domain's key tradeoffs.

    Should I take notes by hand or type them?

    Either works, but keep them in the same tracker you started on Day 1. The value isn't in the notes themselves — it's in being able to see, on Day 27, exactly which five topics you rated lowest across the entire month.

    What to Do If You Fall Behind

    Life happens. If you miss more than 3 days in a row:

    • Don't restart the 30-day clock. Pick up where you left off — the plan is sequenced by domain weight, not by calendar day, so a slower pace still covers the highest-value material first.
    • Compress, don't skip, Week 1. If you're short on time, cut the Anthropic Academy modules to the two most relevant (usually Modules 1 and 3) rather than skipping the hands-on build on Day 6 — the exam rewards practical intuition over passive reading.
    • Never skip the Day 29 full simulation. Everything else in this plan is compressible. A realistic, timed, full-length practice run under exam conditions is the single highest-signal thing you can do before booking the real test.

    Key Takeaways

    • A day-by-day plan beats a vague weekly one because it removes the "what do I actually study today" decision every single time you sit down.
    • Weight your study time to match the exam: Domain 1 (27%) gets a full week; Domain 5 (15%) gets two days.
    • Take the official practice exam twice — cold on Day 7 to find real gaps, and simulated on Day 29 to confirm readiness — not once at the very end.
    • Track your confidence per topic from Day 1. Your Week 4 review is only as good as the weak-spot list you've been building.

    Next Steps

    This plan pairs with structured practice, not just reading. The AI for Anything CCA Practice Test Bank has 200+ scenario-based questions mapped to these same five domains, with explanations for both correct and incorrect answers — useful for the Day 24 and Day 29 checkpoints in this plan if you want more than the official 60 questions to draw from.

    Related reading: How to Pass the CCA-F Exam: Complete Study Guide for the full domain breakdown this plan is built on.
    R

    Rohit Mote

    Founder, AI for Anything

    Rohit Mote is the founder of AI for Anything and builds AI-powered products full-time across the Infinite Products Machine portfolio. Every guide is grounded in hands-on daily use of Claude, Claude Code, and the broader AI tool ecosystem in production systems.

    How we create and review our guides →