Course

Claude Certified Architect — Foundations

A complete route through the CCA-F exam blueprint: the five domains, their published weights, and how the scenario questions are actually written. Fourteen short modules, in the order they build on each other.

14 modules, 6 units, about 183 minutes in total, self-paced

Talk to us Start with a trial

What is in it

Sitting the Exam

How the paper is built and how its questions are written. Six fixed scenarios, four drawn at random, one correct answer against three distractors, 720 to pass on a 100-1000 scale. Start here whatever your background.

  • 1. 15 minutes

    The Claude Certified Architect Exam

    What the paper is, how it is scored, and what 720 out of 1000 actually asks of you.

  • 2. 14 minutes

    Reading a Scenario Question

    How CCA-F questions are written, and how to tell the right answer from the three that are nearly right.

Agentic Systems

The largest domain at 27% of the paper: the loop itself, coordinators and subagents, putting rules in code rather than in prose, and how to break work into sessions.

  • 3. 13 minutes

    The Agentic Loop

    Gather context, take action, verify, repeat: the loop every agentic design is a variation on.

  • 4. 13 minutes

    Coordinators and Subagents

    When to split work across subagents, what a coordinator owns, and what it must never hand off.

  • 5. 12 minutes

    Enforcing Rules in Code

    Why a rule in a prompt is a request and the same rule in code is a guarantee.

  • 6. 13 minutes

    Decomposition and Sessions

    Breaking work into sessions an agent can finish, and knowing where one should end.

Tools and MCP

18% of the paper. What makes a tool description work, the built-in tools, and the Model Context Protocol.

  • 7. 13 minutes

    Tool Design and the Built-in Tools

    What makes a tool description work, and the built-in tools you are expected to reach for first.

  • 8. 13 minutes

    Model Context Protocol

    What the Model Context Protocol standardises, what a server exposes, and when a plain tool is the better answer.

Claude Code

20% of the paper. Where the rules live, planning and refinement, and running it unattended in CI.

  • 9. 13 minutes

    Where the Rules Live

    CLAUDE.md, settings, skills and hooks: which one governs what, and what wins when they disagree.

  • 10. 13 minutes

    Plan Mode and Refinement

    Planning before editing, and the refinement loop that turns a plan into working changes.

  • 11. 12 minutes

    Claude Code in CI

    Running it unattended: permissions, headless mode, and what has to be decided in advance.

Prompting and Structured Output

20% of the paper. Explicit criteria, worked examples, and getting output in a shape a program can rely on.

  • 12. 12 minutes

    Explicit Criteria and Few-Shot Prompting

    Saying what good looks like, and showing it, so output can be judged rather than admired.

  • 13. 13 minutes

    Structured Output and Validation

    Getting output in a shape a program can rely on, and what to do on the run where it is not.

Context and Reliability

15% of the paper, and the one most people meet last: managing the context window, and knowing when an agent should stop.

  • 14. 14 minutes

    Context and Knowing When to Stop

    Managing the context window, and the judgement about when an agent should stop.

The rest of the catalogue

Browse every module, unit and course on the catalogue page, or read how we build and test them on the standards page.