Independent guide
Start here · Step 1

Sign in with your partner-organisation email on the Skilljar site, not claude.ai

Create an Anthropic Academy (Skilljar) account, authenticate, and accept the EULAs — then every course and resource below unlocks.

Important: this is a separate account from claude.ai — not your Claude login. Sign up fresh with your partner-organisation email.

Sign in / sign up at the Anthropic Partner Academy ↗

An independent study guide · Claude Certified Associate - Foundations

Pass the
CCAO-F.

Use Claude well, and know when to escalate. The exact prep path, the exam blueprint, and a guide to all 7 domains.

  1. Sign in at Skilljar with your partner-organisation email — not claude.ai.
  2. Practice on personal gear only — never run Claude on employer devices, networks, or data.
  3. Follow the prep path below (about 6 hrs).
  4. Book the exam — 60 Qs · 120 minutes · closed-book · pass 720/1,000.
60questions
120minutes
720to pass /1,000
7domains
01

Your prep path

This is the official Anthropic Academy prep path for this exam. Complete these courses in order — together they cover every domain. Each course opens on Skilljar — sign in first.

  1. 01 Course 59 min Start ↗
  2. 02 Course 53 min Start ↗
  3. 03 Course 74 min Start ↗
  4. 04 Course 63 min Start ↗
  5. 05 Course 47 min Start ↗
  6. 06 Course 55 min Start ↗
  7. 07 Course 30 min Start ↗
  8. 08 Course 8 min Start ↗

That’s the whole prep. Then practice hands-on (use the domain guides below as reference) and book the exam (sign in with your partner email first). ≈ 6.5 hrs total

02

Exam blueprint

A proctored, closed-book exam: 60 scenario-based multiple-choice / multiple-response questions in 120 minutes, no AI assistance. Scored to 1,000 with a passing bar of 720. It is proctored online or at a Pearson VUE test center, and the credential is valid for 12 months from the date the credential is awarded.

CCAO-F
exam code
60
questions
$99
exam fee
120 minutes
time limit
720/1,000
passing score
12 months
validity

Domains & weights

Study time should follow the weights. Click a domain to jump to its guide.

03

The five domains

Reference for each domain — the concepts to know, the official docs, and exam tips. The prep path above is what to take; this is what to know. Click a domain to expand it.

01

Prompting and Task Execution

14%

Tests whether you can turn a vague business request into a prompt that gets a usable answer quickly, and whether you can break a large or multi-part task into steps Claude can execute reliably.

Key concepts to master
Clear, direct instructions. State the audience, format, and constraints explicitly rather than leaving Claude to infer them; vague requests produce vague, unusable answers on the exam's scenario questions.
Role and context framing. Giving Claude a perspective or role, such as reviewing something as a project manager would, focuses the response; the exam expects you to recognise when framing changes output quality and when it's unnecessary.
Worked examples inside a prompt. Supplying a sample of the desired output format or tone directly in the prompt is one of the most reliable ways to raise quality, and the exam tests whether you reach for it before rewriting from scratch.
Task decomposition. Breaking a large or multi-stage request into an ordered sequence of smaller prompts, rather than one long ask, so each stage's output can feed and be checked before the next begins.
Iterative refinement. Comparing a draft prompt to a revised one and identifying exactly what change improved the output, so refinement is deliberate rather than trial and error.
Structuring long or multi-part requests. Organising a complex ask into labelled parts or sequential steps so Claude addresses every requirement instead of dropping pieces of a compound request.
Matching strategy to task type (analysis, research, drafting, brainstorming). Each task type calls for a different prompting posture, for example open-ended options for brainstorming versus a single defensible recommendation for analysis; the exam expects you to read the task type and adjust.
Output format specification. Telling Claude explicitly whether you need prose, a table, a list, or a specific structure, since an unspecified format is one of the most common reasons a technically correct answer still isn't usable.
Official documentation
Exam pro tips
  • This domain rewards structure over cleverness. Answers that add explicit constraints (audience, format, length) tend to beat answers that just add more adjectives to the request.
  • Task-decomposition questions often contrast "do it all in one prompt" with "break it into steps." Breaking it down is usually correct when stages depend on each other, such as research before drafting.
  • Don't confuse iterating on a prompt with iterating forever. If an output is wrong for a factual reason, no rewording fixes it — that's an evaluation or escalation call, not a prompting one.
  • Scenario questions usually name the task type (research, drafting, brainstorming, analysis) directly in the wording. Treat that word as a clue to the correct strategy, not background color.
02

Output Evaluation and Validation

21%

The single heaviest domain on any Foundations exam. It tests whether you can tell good Claude output from plausible-looking bad output, and what you do about it.

Key concepts to master
Hallucination and confabulation. Claude can generate fluent, specific-sounding details, such as a citation or subsection number, that are simply invented; recognising this failure mode is the foundation of the exam's heaviest domain.
Grounding and citation of sources. Tying a claim back to a specific, checkable source document rather than accepting it because it sounds authoritative or well-formatted.
Acceptance criteria for a task. The specific, concrete standard an output must meet, defined before the task runs so evaluation isn't an afterthought or a matter of taste.
Spot-checking versus exhaustive review. Knowing when reviewing a representative sample of output is sufficient and when the stakes or cost of an error require checking every item.
Human-in-the-loop validation. Building a deliberate human review step into a Claude-assisted process, particularly before high-stakes or externally facing output ships.
Escalation thresholds. The point at which a task's complexity, ambiguity, or risk means the right move is handing it to a subject-matter expert or a Claude Developer or Architect, not iterating on the prompt further.
Reproducibility of a result. Whether re-running a task produces a consistent, comparable answer, which matters when an output is being relied on as a repeatable process rather than a one-off.
Bias and representational harm in output. Recognising when generated content skews unrepresentative, stereotyped, or unfair, which is an evaluation failure distinct from factual inaccuracy.
Official documentation
Exam pro tips
  • This domain is over a fifth of the exam. If you are short on study time, spend it here.
  • Evaluation questions are usually judgment calls, not recall. Expect scenarios where several answers are defensible and one is clearly most responsible.
  • The exam consistently rewards verifying against a source over trusting fluency.
  • Watch for options that fix a symptom by re-prompting when the real answer is to escalate or to validate against a source.
03

Product and Model Selection

12%

Tests whether you pick the right Claude surface for a task, whether chat, a Project, research mode, or an artifact, and the right model tier for the task's stakes, while managing context sensibly across a working session.

Key concepts to master
Claude.ai product surfaces: chat, Projects, Artifacts, research mode. The distinct places you can work in Claude, each suited to different tasks: a one-off chat, a persistent Project, a rendered Artifact, or a research-mode session. The exam tests picking the right one for the job.
Model tiers: Haiku, Sonnet, Opus. Anthropic's model lineup trades off speed, cost, and reasoning depth; knowing the shape of that trade-off matters more for the exam than memorising version numbers.
Cost, speed, and quality trade-offs. The core selection logic behind every model or feature choice: matching how much power a task actually needs against what it costs in time and money to get it.
Context window. The amount of text and history Claude can hold in a single conversation or Project at once; understanding its limits explains why long sessions or huge uploads degrade.
Context degradation over long conversations. The way answer quality can drop as a conversation grows very long, losing track of earlier detail even though the context window hasn't technically been exceeded.
When to restart a conversation versus continue it. Recognising the signs a conversation has outgrown its usefulness, such as repetition, drift, or degraded answers, and knowing to start fresh with a carried-forward summary rather than pushing on.
Summarizing or persisting context. Deciding whether information should be condensed and carried into a new conversation or stored permanently in a Project's knowledge, versus left in a single disposable chat.
Matching a tool to a task. The general discipline behind this whole domain: choosing the product surface and model that fit a task's actual requirements, not the most capable or most familiar option by default.
Official documentation
Exam pro tips
  • "Always pick the most powerful model" is a trap answer. The exam tests cost/speed/quality trade-off awareness, and for high-volume, low-complexity tasks the fast, cheap model is the correct choice, not the cautious one.
  • Expect at least one question where the fix for a degraded long-running conversation is starting a new conversation with a carried-forward summary, not pushing the same thread further.
  • Know structurally what a Project is for, meaning persistent instructions and knowledge across many chats, versus a single chat with no persistent memory. Questions often hinge on that distinction alone.
  • Don't over-index on exact model names or version numbers. The guide tests the selection logic, cost, speed, quality, and task fit, not a spec sheet.
04

Workflow Integration and Solution Design

16%

Tests whether you can place Claude correctly inside a real business process: spotting where it helps, designing how it fits, and explaining its value and its limits to people who weren't in the room when you built it.

Key concepts to master
Requirements and use-case analysis. Identifying what a workflow actually needs before proposing Claude as part of the solution, so the fit is based on the task rather than the tool.
Current-state versus future-state process mapping. Documenting how a process runs today versus how it would run with Claude involved, which is what makes a proposed change concrete enough to evaluate.
Augmenting a workflow versus redesigning it. Augmenting means Claude assists an existing step without changing the process; redesigning means the process itself changes because Claude is involved. The exam expects you to tell these apart.
Solution design and iteration. Treating a first Claude-supported workflow as a draft to be tested and revised, not a finished deliverable the moment it works once.
Piloting before full rollout. Scoping a small, contained trial of a new Claude-supported workflow before proposing it organization-wide, so problems surface at low cost.
Stakeholder communication. Explaining what a workflow change involves and why to the people affected by or approving it, in terms they can act on.
Setting realistic expectations about value and limits. Describing honestly what Claude will and won't reliably do for a given workflow, rather than overselling capability or glossing over where human review stays essential.
Change management for AI-enabled workflows. Managing the human side of a workflow change, such as training, feedback, and adoption, not just the technical configuration, since a redesigned process still has to be adopted by people.
Official documentation
Exam pro tips
  • This domain is scenario-heavy. Expect a paragraph describing a messy real workflow and a question about where Claude fits, not a definitions question.
  • The exam distinguishes "helps with a task inside the workflow" from "replaces the workflow." Full automation is rarely correct when the scenario has judgment calls, exceptions, or compliance steps in it.
  • When a question asks how to communicate Claude's value to a stakeholder, answers that overstate reliability or omit limitations are the distractors. Honest scoping is the credential's whole ethos.
  • Don't skip this domain because it feels soft. At 16% it outweighs Product and Model Selection and Configuration and Knowledge Management individually.
05

Configuration and Knowledge Management

12%

Tests whether you can set up a Claude Project properly, with instructions, knowledge sources, and connectors, and keep that configuration accurate as things change.

Key concepts to master
Project custom instructions. The standing directions a Project applies to every chat inside it, covering role, tone, constraints, and output format, so you don't restate them each time.
Project knowledge base and uploaded files. The documents attached to a Project that Claude can draw on directly; curating what's in it is as much a part of configuration as writing the instructions.
Connectors, for example Google Drive and Gmail. A live link to an external source that keeps a Project's context current automatically, as distinct from a static one-time file upload.
System-level instructions. Instructions specific and durable enough to consistently shape behavior across many interactions, rather than a vague tone request that only sometimes takes effect.
Knowledge curation and pruning. Actively removing outdated, redundant, or irrelevant material from a Project's knowledge base, since excess content can crowd out what's actually relevant and degrade answers.
Configuration drift and maintenance. The way a Project's instructions and knowledge fall out of date as the underlying process changes, unless someone deliberately reviews and updates them.
Access and sharing within a Project. Who can see, use, or edit a Project's configuration and knowledge, which matters as much for governance as for collaboration.
Updating instructions as processes change. Treating a Project's setup as a living configuration to revisit when the underlying task or policy changes, not a one-time setup step.
Official documentation
Exam pro tips
  • Questions here often hinge on maintenance, not setup. A Project configured once and never revisited is the wrong-answer pattern, since the domain explicitly covers informing, maintaining, and updating configurations.
  • Don't assume more uploaded knowledge is always better. Irrelevant or stale files can crowd out the documents that matter and degrade answers; curation is part of the job.
  • Connectors and one-time uploads are tested as a real choice, not interchangeable options. Know which one keeps information current automatically.
  • This is one of the two domains, alongside Governance, most likely to test who should have access to a configuration, not just what should be in it.
06

Governance, Risk, and Responsible Use

15%

Tests your judgment about when Claude should and shouldn't be used, how to handle sensitive data responsibly, and whether you follow your organization's AI policy rather than your own preference.

Key concepts to master
Appropriate versus inappropriate use cases. Some tasks are a poor or unacceptable fit for AI assistance regardless of how carefully they're handled; recognising these is distinct from handling a permitted task carefully.
Data sensitivity and classification. Recognising which categories of information, such as regulated personal data, confidential business material, or credentials, require safeguards before they go anywhere near a prompt or upload.
Regulatory and privacy considerations. Legal and policy constraints on how data can be used or shared, which apply to AI-assisted work exactly as they apply to any other handling of that data.
Organizational AI policy and governance standards. Your employer's specific rules for AI use, which take precedence over general good judgment when the two would otherwise conflict.
Ethical implications of AI use, including bias and fairness. Whether a use of Claude is appropriate goes beyond whether it's permitted; the exam tests recognising bias, overreliance, and fairness concerns as real considerations, not just policy compliance.
Data minimization and anonymization. Removing or masking sensitive identifiers before data is used, which is what actually satisfies most privacy policies, as opposed to merely instructing a tool not to retain the data.
Accountability for AI-assisted output. The person using Claude remains responsible for the accuracy and appropriateness of the resulting work, not the tool itself.
Escalating policy-ambiguous situations. Recognising when a scenario isn't clearly covered by existing policy and needs a decision from someone with the authority to make it, rather than guessing.
Official documentation
Exam pro tips
  • This is the second-heaviest domain after Output Evaluation. Treat it as core material, not a compliance afterthought.
  • The recurring correct pattern in governance scenarios is redact or anonymize, then proceed, not proceed as-is and not refuse the task outright. Both extremes are usually the distractors.
  • Telling Claude not to retain data is not a substitute for removing sensitive data before you share it. The exam tests this distinction directly.
  • Watch for scenarios that are technically permitted but still a bad idea. The ethical-implications objective means "is this allowed" and "is this appropriate" are tested as separate questions.
07

Troubleshooting and Optimization

10%

Tests whether you can diagnose why a prompt or workflow is underperforming, correct course based on results, and tune a working process for efficiency rather than leaving it as first shipped.

Key concepts to master
Root-cause diagnosis of poor output. Tracing a bad result back to its actual cause, such as ambiguous instructions, missing context, the wrong model, or stale knowledge, instead of guessing at a fix.
Common failure modes: ambiguous instructions, missing context, wrong model choice, stale knowledge. The recurring, recognisable reasons a Claude-assisted task underperforms; naming the correct one is most of what a troubleshooting question is testing.
Feedback-driven adjustment. Using specific feedback on an output, such as too long, wrong tone, or a missed requirement, to make a targeted correction rather than a full rewrite.
Workflow efficiency versus effectiveness. Efficiency problems mean a task takes too long or too many steps; effectiveness problems mean the output quality itself is wrong. The fixes for each differ, so diagnosing which one you have comes first.
Regression after a change. A fix or adjustment that solves one problem but introduces or reintroduces another, which is why changes should be checked against the whole task, not just the symptom they targeted.
Isolating variables when testing a fix. Changing one thing at a time when troubleshooting a prompt or workflow, so you can tell what actually caused the improvement.
Continuous improvement mindset. Revisiting a workflow that already works to see whether it could be simpler, faster, or less manual, rather than leaving it untouched once it functions.
Recognising when the fix isn't a prompt fix. Some problems, such as a genuinely unsuited task, missing source data, or a policy conflict, won't be solved by rewording, and the exam rewards knowing when to stop iterating on the prompt.
Official documentation
Exam pro tips
  • This is the lightest domain by weight, but don't skip it. Its questions overlap conceptually with Prompting and Task Execution and with Output Evaluation, so studying it reinforces both.
  • The exam separates "the output is wrong," an evaluation issue, from "the process to get the output is inefficient," a troubleshooting issue. Read what's actually being asked before picking a fix.
  • A recurring trap is re-running the same prompt with minor wording changes when the real problem is missing context or a mismatched model. Diagnose before you iterate.
  • Optimization questions often reward removing steps or consolidating a workflow, not adding more prompts or more review layers.
04

Lab & study setup

The Associate exam tests applied judgment, not architecture, so preparation should center on real, low-stakes Claude usage rather than a technical lab. Build an actual Project, feed it real but non-sensitive knowledge, and put each domain's habits into practice on genuine business tasks.

⚠ Read this first: practice on personal accounts and non-production data

Do all exam preparation on a personal Claude account, using your own data or clearly non-sensitive material. Never point practice exercises at production systems, customer data, an employer's network, or anything a workplace policy would classify as confidential or regulated. This is basic professional hygiene, not a rule specific to any employer: keep your study environment fully separate from any production or corporate system.

  • Use a personal Claude account for practice, separate from any account or seat an employer manages.
  • Build a practice Project's knowledge base from sample, public, or deliberately fabricated documents, never real customer records, regulated personal data, or confidential company material.
  • If you want to practice on work-relevant scenarios, recreate them with fictional names and numbers rather than uploading the real thing.
  • Check your own organization's AI usage policy before using any work-related tool or account for exam practice, and default to personal tools when in doubt.
Subscriptions & access
Your environment
  • A Claude.ai account on a plan that supports Projects and Artifacts; the free tier is limited, so a paid plan gives more room to practice realistically.
  • One dedicated practice Project, so configuration work doesn't get mixed into everyday chat history.
  • A small folder of sample business documents you're free to use however you like, such as a fake policy memo, a mock spreadsheet, or a sample report, with enough variety to practice knowledge curation.
  • Access to a connector you're willing to practice with, such as a personal Google Drive or Gmail account, if you want hands-on reps configuring connectors rather than only uploads.
Hands-on practice
  • Prompting and Task Execution: take one real or realistic multi-step task, such as preparing a status update from scattered notes, and run it once as a single detailed prompt and once as a chained, step-by-step version. Compare the results.
  • Output Evaluation and Validation: give Claude a task with a checkable answer, such as summarizing a document you already know well, and deliberately hunt for anything it got wrong, invented, or overstated before calling the output done. Also practice choosing between an inline answer, a structured table, and a full artifact for the same task, and notice when each format actually serves the reader better.
  • Product and Model Selection: if your plan allows it, run the same task on different models and notice the difference in speed and quality; practice deciding when a quick chat is enough versus when a task belongs in a Project.
  • Workflow Integration and Solution Design: pick one recurring task from your own work, or a plausible one, and sketch how Claude would fit into it end to end, including where a human still needs to check the output.
  • Configuration and Knowledge Management: build out a Project's custom instructions and knowledge base from scratch, then revisit it a week later and update or prune it as if the underlying process had changed.
  • Governance, Risk, and Responsible Use: practice spotting what should never go into a prompt or upload, and rehearse anonymizing a realistic-looking document before using it as sample knowledge.
  • Troubleshooting and Optimization: take a prompt that produces a mediocre result and diagnose why before rewriting it. Is it missing context, the wrong task type, or a genuinely bad fit for Claude?
From people who passed
  • This exam rewards judgment built from real use, not memorized definitions. A few weeks of actually running Projects and evaluating output serves you better than re-reading the blueprint.
  • Spend real time on Output Evaluation and Validation and on Governance, Risk, and Responsible Use. Together they're over a third of the exam, and both are scenario-judgment domains that don't reward guessing.
  • Don't prepare only by chatting. Build at least one real Project with instructions and uploaded knowledge; several domains assume you've actually done this, not just read about it.
  • If your day job doesn't give you hands-on Claude time, manufacture some: pick a personal task, such as planning something or organizing information, and run it through Claude deliberately, evaluating each output as if it had to go to someone else.

Sources:

05

Official resources

There is no single official “CCAO-F study guide” PDF. The real prep is the Anthropic Academy courses plus the official product docs — everything below is official.

  • Book the exam ↗

    The certification page where you schedule and sit the proctored exam. Sign in first — it's partner-gated, so the page only appears once you're signed in. — CCAO-F certification page

  • Partner portal — exam access ↗

    Where you actually book the exam. Sign in with your partner-organisation account to request the Claude Certified Associate – Foundations exam. — claude.com/partners

  • Anthropic Academy — prep path ↗

    The free, self-paced course path this guide's prep section is built from — sign in and enroll. — anthropic-partners.skilljar.com

  • All Claude certifications ↗

    The index of all four Claude certifications, with links to every certification page and exam guide. — Partner Academy certifications page

  • Claude documentation ↗

    Anthropic's official documentation — Claude's capabilities, Projects, and safe, effective everyday use. — docs.claude.com

⚠ A word of caution

Skip third-party “free CCAO-F practice test” and brain-dump sites that crowd search results. They are not affiliated with Anthropic, are frequently inaccurate, and dump sites can violate exam terms. Stick to Anthropic Academy and the official docs above.

06

Frequently asked questions

The essentials on eligibility, cost, format, and access — answered from verified facts.

Who is the Claude Certified Associate – Foundations exam for?

Professionals who use Claude as a day-to-day productivity tool rather than build with it: operations, marketing, project management, education, communications, and general knowledge-work roles. The exam guide places the audience between casual chat users and technical AI practitioners, expecting candidates to build and maintain real Claude Projects, not just ask one-off questions.

Do I need to know how to code to pass?

No. The certification is explicitly non-technical: no software-development or API experience is needed or tested. Everything is scoped to Claude's standard product surfaces, such as chat, Projects, and Artifacts, rather than building against Claude.

What does the exam cost?

$99 USD for the proctored exam itself, payable at registration through the Anthropic Partner Academy. Any discount shown at checkout reflects your partner tier.

Are the official prep courses free?

Yes. The Associate prep courses on the Anthropic Partner Academy are free to enroll in and work through. The $99 fee applies specifically to registering for the proctored exam, not to preparation.

How many questions are on the exam, and how much time do I get?

60 multiple-choice and multiple-response items in 120 minutes. Each item states how many responses to select.

What score do I need to pass?

A scaled score of 720 on a 100–1,000 scale. It's criterion-referenced: you're measured against a fixed performance standard, not ranked against other candidates. Your score report also shows percent-correct by domain, but that breakdown is informational only; your pass/fail result is based on the total scaled score.

How long does the credential stay valid?

12 months from the date it's awarded. Because Claude and its surrounding products change quickly, the credential is deliberately time-limited. On-time renewal is a free, non-proctored assessment on the Anthropic Partner Academy; a lapsed credential requires retaking the full exam at full fee.

How does this differ from the Claude Certified Developer credential?

Developer is the technical counterpart: it assumes hands-on software-engineering experience and tests API mechanics, agent construction, and tool and integration work, for a higher fee ($125) and a shorter exam (53 items). Associate stays entirely inside claude.ai's product surfaces and tests judgment and workflow use, not implementation.

How does this differ from the Claude Certified Architect credentials?

Architect – Foundations and Architect – Professional test designing and operating Claude systems at an organizational or enterprise level: architecture patterns, integration, and lifecycle and stakeholder management for people who own that design. Associate is scoped to using Claude effectively inside your own role and recognizing when to escalate harder technical or architectural work to a Developer or Architect, not to designing the system yourself.

What happens if I don't pass?

You can retake the exam after a waiting period that increases with each attempt: 14 days after a first fail, 30 after a second, 90 after a third, up to four attempts total within a rolling 12-month period. The exam fee applies again on each retake.

Do I need any prior certification or experience to sit this exam?

No. There are no mandatory prerequisites or required courses for this or any Claude certification. Regular, hands-on experience using Claude in a professional setting is recommended, not required; the credential is awarded on exam performance alone.