learnclaude.today CCAR-F exam prep · 60 questions
Request an invitation
60 questions · 120 minutes · pass at 720/1000

Walk into CCAR-F already knowing what the questions look like.

A self-paced guide to all five exam domains — 61 lessons, 91 practice questions, three full 60-question mocks weighted exactly like the real paper, and a study assistant that answers out of the course instead of guessing.

Request an invitation Invitations go out as soon as possible. Nothing is charged today — the price below is what it will cost when access opens.
  • 61 lessons · 6h 40m
  • 91 practice questions + 3 mock exams
  • Study AI assistant included
  • 2h 40m narrated
A CCAR-F lesson open in the guide: the course index on the left, the stop_reason table, a key-idea callout, and one of the reader's own highlights.

The guide itself — every screenshot on this page is the real product, not a mockup.

What's inside

All five domains, at the weights the exam actually uses.

The blueprint is public, so the course is built to it: every domain gets lesson time and question count in proportion to how much of the paper it is. Nothing is padded to look thorough.

61
lessons · 23 sections

Plus a 9-lesson orientation module — what the paper tests, how it is scored, and the one mental model that carries half of D1.

Simulators

Three full mock exams, and a drill after every domain.

A module simulator after every domain — ten questions drawn from that domain's pool, in the real format, marked immediately. Then three separate 60-question forms, 120 minutes each, distributed 16 / 11 / 12 / 12 / 9 across the domains by their true weight and scaled to 1000 with the pass line at 720. Their 180 questions are written for the mocks alone and appear nowhere else in the course, so a form measures what you know rather than what you remember answering.

  • A real clock, and a five-minute warning you cannot miss.
  • Single-answer and multiple-response, each item stating how many to select.
  • Close the tab mid-attempt and it is still there when you come back.
  • Every attempt kept, per form, so you can see the scaled score moving.
  • Six production scenarios per form, several decisions inside each — the way the paper asks.
Simulator A: 60 questions across D1–D5 at 16/11/12/12/9, a best of 783 over two attempts, with Simulator B beginning below and all three listed in the index. An attempt in progress: 60 questions, 11 answered, 119:37 on the clock, a seven-question scenario set open, and a multiple-response item reading Select THREE.
A graded question showing the correct option marked Correct, a full explanation of why the distractors fail, and an Ask AI about this question button.
Feedback

A wrong answer is where the studying actually happens.

Every question carries a written explanation that names each distractor by what it says — never "as in option B", because the options are shuffled on every render. Multi-selects tell you which correct answer you missed, not just which box was wrong.

Still not convinced? One click hands that exact question, your answer and the explanation to the study assistant and asks it to take the argument further.

Study assistant

It answers out of the course, not out of a memory of the docs.

Ask anything and the guide searches all 61 lessons first, then answers with those passages in hand — and shows you which lessons it found, as chips that jump straight there. It is included in the price: no API key, no separate subscription.

  • Attach a lesson, a diagram, a graded question or one of your own notes and ask about that.
  • Drag across any sentence and ask about the sentence.
  • Keep an answer: save it back into your notes, filed under the lesson it came from.
The Ask AI panel answering a question about stop_reason, with Related lessons chips underneath and a Save as note action.
The notes panel: highlights grouped by domain, each with the quote, the reader's note and their own tags, filtered by colour and tag.
Notes & highlights

Highlight it, tag it, and actually find it again in week four.

Four highlighter colours, a note on any passage, and your own tags — exam, shaky, whatever you use. The panel lists every one in course order, filtered by colour, tag or page, and clicking a row scrolls to the exact sentence.

It all lives on your account, so the highlight you made on your laptop is on your phone. And resetting your progress never touches your notes — they are your writing, not a score.

Built to be lived in

Six months is a long time to spend in a badly-built page.

The same guide in the other theme: an MCP host-client-server diagram, re-rendered in that palette.
Light and dark, properly
Follows your system by default. Diagrams re-render in the palette you are actually wearing.
Your text size
90% to 150%, without the layout falling over — and it stays where you put it.
Keyboard, if you want it
M to mark complete, H to highlight, ? for the rest. Switchable off.
2h 40m narrated
All 61 lessons recorded, with the paragraph you are hearing lit up as it plays.
The guide on a phone: the header collapses to CCAR-F, the lesson reads full width, highlights intact.

Revising on a phone is not an afterthought

Nothing is clipped, nothing scrolls sideways, and the annotation toolbar comes up on a long press the way it should — so a highlight you make on the train is waiting on your laptop.

Progress, notes and exam attempts sync to your account across every device you sign in on.

Optional hands-on track

Then prove it in code, offline, with no API key.

Five Go labs, one per domain, that ship red: 63 failing tests are your worklist. The agent loop, tool errors and tool_choice, Claude Code config resolution, structured output, escalation and compaction — graded against replayed API responses, so nothing calls the network and nothing costs you a token.

Needs a Go toolchain. Everything else in the course runs in a browser.

$ cd lab/d1 && go test ./...
FAIL  7 tests
 
# your worklist for D1. when they are green,
# you have written the agentic loop yourself.
 
d1 · 7d2 · 10d3 · 16d4 · 13d5 · 17
Pricing

One payment, six months, everything above.

Long enough to read it properly, sit the mock more than once, and still have room for the retake nobody plans for.

Invitation only, for now
Payments are not open yet. Ask for an invitation and you pay nothing until they are.
Full access
$99 USD, once
Six months from the day you sign in. No renewal, no card kept on file.
  • All 61 lessons across the five domains, plus orientation
  • 91 practice questions in the per-domain simulators
  • 180 more written for the mocks alone, and used nowhere else
  • Three 60-question, 120-minute weighted mocks — retake any of them as often as you like
  • The study assistant, included — no API key of your own
  • 2h 40m of narration, notes, highlights and cross-device sync
  • The five hands-on Go labs
Request an invitation
Nothing is charged today — this asks for an invitation, it does not buy anything.
The exam fee is not included
CCAR-F itself costs US$125, paid to the testing vendor when you book, and the certification is valid for 12 months. This is the preparation, not the sitting.
Every fact is sourced
Written against the published exam blueprint and Anthropic's own documentation, with a provenance record behind each claim. Where the docs and the course ever disagree, the docs win and the course gets fixed.
Nothing to install
It is a web page you sign in to, at learnclaude.today. The labs are the one exception — those want Go on your own machine.
After six months
Access ends and nothing renews — no card is kept on file. You can buy another six months at the price of the day.

The paper is 60 questions and two hours. Go in having already done it.

$99 once, six months of access, and three mock exams waiting for you whenever you want to know where you stand. Invitation-only while we open up.

Request an invitation I already have access