Claude Certified Architect (Foundations): Exam Guide and How I Passed

The Claude Certified Architect (Foundations) exam, code CCAR-F, is Anthropic's certification for solution architects who design and build production applications with Claude. It tests Claude Code, the Claude Agent SDK, the Claude API and the Model Context Protocol, and it is open to people at organisations in the Claude Partner Network. This guide sets out the exam facts, what the exam covers, a four-week study plan built from the resources below, and a first-hand account of sitting it.

I took the exam at the start of the week and had my result by midweek, I passed. Below are the resources I leaned on and how the exam actually went.

One caveat up front. Some of these resources are official and some are community-made, and I've flagged which is which. The community ones are genuinely useful, but treat them as a study aid rather than gospel. A bit of vibe code energy, helpful but not authoritative.

What is the Claude Certified Architect (Foundations) exam?#

• Provider: Anthropic's Claude Certification Program. You register and pay on Anthropic Partner Academy, and Pearson VUE delivers the exam.
• Who can take it: any individual at a partner organisation in the Claude Partner Network.
• Delivery and proctoring: online proctored through Pearson OnVUE, or at a Pearson test centre. Pearson replaced the previous ProctorFree setup on 30 June 2026.
• Questions: 60 multiple-choice and multiple-response items, each stating how many responses to select, framed by 4 scenarios drawn at random from a bank of 6.
• Duration: 120 minutes. Anthropic suggests planning for about 135 minutes at your seat, to cover check-in and a short survey.
• Passing score: a scaled score of 720 on a scale of 100 to 1,000. You see your score on screen at the end of the exam.
• Cost: $125 USD. A retake costs the full exam fee.
• Validity and renewal: 12 months from the date you earn it. On-time renewal is free, through a non-proctored assessment on Anthropic Partner Academy; a lapsed certification means retaking the full exam.
• Retakes: waiting periods of 14 days after a first failed attempt, 30 days after a second and 90 days after a third, with up to 4 attempts in any rolling 12 months.
• Language: English only.

These facts come from the exam guide and certification FAQ on Anthropic's certification page, and from Pearson VUE's Anthropic page. The proctoring setup, practice exam and results timing in the sections below are from June 2026, before the move to Pearson. Anthropic's FAQ now states that there is no practice exam, and that the exam guide includes sample questions instead.

What does the Claude Certified Architect exam cover?#

The exam guide splits the exam into five content domains, weighted by their share of scored items. Agentic architecture carries the most weight and context management the least.

Every question sits inside a scenario drawn from realistic customer use cases. The six scenarios in the guide are a customer support resolution agent, code generation with Claude Code, a multi-agent research system, developer productivity with Claude, Claude Code for continuous integration, and structured data extraction. Two of them are about agentic coding with Claude Code, which is one reason the CI/CD material below matters.

How to prepare: a four-week study plan#

This plan uses only the resources listed in the rest of this guide. It follows the order recommended further down, with the official material first and the community resources after.

• Week 1: the exam guide and the API. Read the exam guide in full, including the sample questions, and note which of the five domains you know least. Then take Building with the Claude API and Introduction to Model Context Protocol on Skilljar.
• Week 2: Claude Code and skills. Take Claude Code in Action and Introduction to Agent Skills. Set up Claude Code on a real project as you go, because the guide's preparation exercises expect you to have configured it yourself.
• Week 3: patterns in real code. Work through several of the claude-cookbooks and compare them with the courses. Spend time on the panaversity CI/CD material, and use the CCA-F study material site to check which concepts to know and which traps to avoid.
• Week 4: practice and review. Read the paullarionov guide and repeat its mock test until your weak domains improve. Finish by rereading the exam guide's sample questions, slowly, looking for the qualifier each answer turns on.

The plan assumes you already build with Claude: the exam guide expects about six months of hands-on experience with the API, Agent SDK, Claude Code and MCP. For teams, re:cinq runs Claude Code and agentic engineering training.

Start here: the official docs#

Read these before touching anything else. The point of reading them first is to know what to skip later.

The Exam Guide (official). When I sat the exam, you got this once you were granted access; Anthropic now publishes the Foundations exam guide openly. It describes the exam content, lists the domains and task statements that get tested, includes sample questions, and recommends how to prepare, which together tell you what to study and what to skip. Reading it first saved me from over-studying corners that never came up.

The FAQ (official). When I sat it, a practice score of around 900 was a good sign I was ready. The practice exam has since been withdrawn, so use the official exam guides' sample questions instead.

The official courses#

Anthropic publishes free training on Skilljar, and four courses cover what the Foundations exam actually tests:

• Building with the Claude API
• Claude Code in Action
• Introduction to Agent Skills
• Introduction to Model Context Protocol

Work through these before the community guides, since the exam maps closely to this material.

Community study resources#

paullarionov/claude-certified-architect guide. This one was a genuinely useful read. It walks through the concepts in a way that helped me make better choices on the mock exam. The repo also includes a PDF version of the guide and a mock test you can repeat as many times as you want, which is where most of my practice reps came from besides the official mock exam.

The CCA-F study material site. Good for the "know this, avoid that" framing. It points you at concepts worth knowing and flags common traps. It also links out to the suggested study materials, which are worth reading rather than skimming.

For grounding the concepts in real code#

Anthropic's claude-cookbooks. These are worked examples of how the concepts are actually implemented. I read through several of them and compared the code against what the courses taught. Even reading a handful made the abstract ideas concrete, which helped on the questions that test whether you actually understand a pattern rather than just recognize its name.

Claude Code in CI/CD pipelines (panaversity). I'm calling this out separately because the real exam had quite a few questions on git pipeline commands, more than I expected from the mock. This page has good examples of that material. If git and CI/CD aren't part of your daily muscle memory, spend time here.

ProctorFree exam setup#

When I sat the exam in June 2026, it ran online through a tool called ProctorFree, and getting from the invite to the first question took more steps than I expected. An exam email arrives with a link, and following it walks you through downloading ProctorFree, installing it, and clearing its pre-exam setup, which runs a hardware and connection check across your camera, microphone, network, and screen recording. Once that passes you begin, and the whole session is recorded from start to finish. The step that's easy to fumble is the very end: after you submit, leave everything open until the recording has finished uploading.

A few things made the setup smoother for me. Run on a single screen and unplug any external monitors before you start, then shut down anything you don't need open. The obvious culprits are chat apps like Slack, Teams, and Discord, and the proctor also flags less obvious ones:

• other AI assistants, like Claude or ChatGPT
• translation tools
• IDEs and note-taking apps
• any stray browser tabs

Clear your desk, check that your camera and mic are working, and you should be ready.

At the end of the exam, after you submit, wait until the recording upload has fully completed before closing anything or shutting down.

The proctoring wasn't the hard part of the day, but it's worth getting your environment right beforehand so it never becomes one. Since 30 June the exam runs on Pearson VUE instead, so check Pearson's own requirements before you sit it.

How the exam actually went#

When I sat it, scoring was officially 5 or more days, but mine came back noticeably faster than that. Since the move to Pearson, you see your score on screen at the end of the exam.

The biggest gap between the mock and the real exam was the wording, not the difficulty of the underlying concepts. The real questions used more complex sentences, and they hide key qualifiers inside them. The correct answer often turns on a single word you have to slow down and find. On the mock I could move fast, but on the real exam reading carefully was the whole game.

What I'd tell you before you sit it#

• Read the official Exam Guide first, so you know what to skip but also in which direction to go when answering. This is the single most useful thing you can do.
• When I sat it, the practice exam was my readiness gauge, and I didn't book until I was clearing 900/1000 comfortably. It has since been withdrawn, so work through the official exam guide's sample questions instead.
• Slow down on the wording. The real questions bury qualifiers in longer sentences. Find the qualifier before you pick.
• Don't skip git and CI/CD. The pipeline command questions showed up more than the mock suggested, and the panaversity page is a good drill.
• Read a few cookbooks, not just course slides. Seeing the patterns in code is what separates recognizing a concept from understanding it.
• Build the exercises the guide describes and actually run them, which pays off most if your experience so far leans more toward reading about the patterns than writing them, because working through something you can run and break yourself sticks far better than reading the description and nodding along.
• Take the community resources seriously but lightly. They're a great head start, not the source of truth. The official guide and FAQ are.

That's the lot. If you're on the fence about sitting it, the prep is doable with the resources above, and the certification is a solid checkpoint for getting your fundamentals straight. If your team is preparing together, our AI education programme covers the same material with hands-on work on your own codebase.

Which Claude certifications exist and where Foundations fits#

Pearson VUE

lists four Claude certifications, each with its own exam guide on Anthropic Partner Academy. Certification is open to organisations in the Claude Partner Network.

Architect Foundations covers the core building blocks: Claude Code, the Agent SDK, the API and MCP. Architect Professional extends the architect role into enterprise integration and governance. The Professional exam guide states no mandatory prerequisites, so Foundations is a sensible first step rather than a required one. Pearson also notes that the Associate, Developer and Architect Professional exams change on 7 December 2026, with updated guides from 17 November 2026.