Associate, Developer, Architect Foundations and Architect Professional - every track, not just the entry one. One question from every exam domain, in the browser, with no account.
Written from scratch against the official exam guides and weighted to the published blueprint. Every answer option gets an explanation of why it is - or isn't - correct.
Each page runs one question from every domain of that exam - 5 to 8 questions, covering the whole official blueprint. Nothing is stored, nothing is asked for.
One question from the Associate bank, worked through in full - so you can judge the quality before you click anything.
A project manager is configuring a Claude Project to help employees ask questions about travel, expense, and leave policies. Most requests are repetitive, such as identifying which policy applies or drafting a standard answer from approved guidance, but some involve exceptions, conflicting policy language, or situations that may require HR or compliance judgment. The team wants fast responses and reasonable cost without letting Claude make unsupported policy decisions. Which model-selection approach is most appropriate?
Choose whichever model allows the project to fully automate final policy decisions so managers no longer need to review escalated cases.
Model selection does not eliminate the need for escalation or human judgment. Policy decisions that are ambiguous, consequential, or compliance-sensitive still require human review even when Claude assists with analysis.
Use Opus for every policy question so that all employees receive the highest-quality answer regardless of whether the request is simple or complex.
Using the most capable model for all requests ignores cost and latency trade-offs. Straightforward, repetitive policy questions do not usually justify the expense and slower response time of the highest-capability model.
Use Haiku for both routine triage and employee exception requests because faster responses are more important than deeper interpretation in a self-service policy tool.
Haiku is well suited to fast, lower-cost handling of simple tasks, but nuanced exceptions may require more careful reasoning and review. Prioritizing speed alone can increase risk when answers affect employee rights, compliance, or approvals.
Use Haiku for high-volume triage and simple draft replies where the policy rules are clear, and route ambiguous or higher-risk cases to a more capable model such as Sonnet or Opus with human review.
This matches model capability to task complexity, volume, cost, and risk. Routine classification can be handled efficiently, while nuanced policy interpretation receives stronger reasoning support and appropriate human oversight.
| Certification | Items | Time | Pass | Valid | Domains |
|---|---|---|---|---|---|
| Architect - Foundations | 60 | 120 min | 720 | 12 months | 5 |
| Architect - Professional | 63 | 120 min | 720 | 12 months | 7 |
| Associate - Foundations | 60 | 120 min | 720 | 12 months | 7 |
| Developer - Foundations | 53 | 120 min | 720 | 12 months | 8 |
All four are scored on the same scale: pass at 720 out of 100-1,000, with percent-correct by domain on the score report.
| Exam code | CCAR-F |
|---|---|
| Number of items | 60 |
| Item format | Multiple-choice and multiple-response; each item states how many responses to select |
| Exam structure | 4 scenarios drawn from a bank of 6 |
| Time limit | 120 minutes |
| Passing score | Scaled score of 720 on a scale of 100-1,000 |
| Validity period | 12 months from the date the credential is awarded |
| Delivery | Proctored: online proctored and/or test center |
| Domains | 5 |
Official blueprint weighting beside the number of questions this site holds for each domain, so you can check the bank follows the blueprint rather than take it on trust.
| Domain | Official weight | Questions here |
|---|---|---|
| Agentic Architecture & Orchestration | 27% | 96 |
| Tool Design & MCP Integration | 18% | 66 |
| Claude Code Configuration & Workflows | 20% | 72 |
| Prompt Engineering & Structured Output | 20% | 72 |
| Context Management & Reliability | 15% | 54 |
| Total | 100% | 360 |
| Exam code | CCAR-P |
|---|---|
| Number of items | 63 |
| Item format | Multiple-choice and multiple-response; each item states how many responses to select |
| Time limit | 120 minutes |
| Passing score | Scaled score of 720 on a scale of 100-1,000 |
| Validity period | 12 months from the date the credential is awarded |
| Delivery | Proctored: online proctored and/or test center |
| Domains | 7 |
Official blueprint weighting beside the number of questions this site holds for each domain, so you can check the bank follows the blueprint rather than take it on trust.
| Domain | Official weight | Questions here |
|---|---|---|
| Solution Design & Architecture | 17% | 33 |
| Claude Models, Prompting & Context Engineering | 13% | 24 |
| Integration | 19% | 36 |
| Evaluation, Testing & Optimization | 16% | 30 |
| Governance, Safety & Risk Management | 14% | 27 |
| Stakeholder Communication & Lifecycle Management | 14% | 27 |
| Developer Productivity & Operational Enablement | 7% | 12 |
| Total | 100% | 189 |
| Exam code | CCAO-F |
|---|---|
| Number of items | 60 |
| Item format | Multiple-choice and multiple-response; each item states how many responses to select |
| Time limit | 120 minutes |
| Passing score | Scaled score of 720 on a scale of 100-1,000 |
| Validity period | 12 months from the date the credential is awarded |
| Delivery | Proctored: online proctored and/or test center |
| Domains | 7 |
Official blueprint weighting beside the number of questions this site holds for each domain, so you can check the bank follows the blueprint rather than take it on trust.
| Domain | Official weight | Questions here |
|---|---|---|
| Prompting and Task Execution | 14% | 48 |
| Output Evaluation and Validation | 21% | 78 |
| Product and Model Selection | 12% | 42 |
| Workflow Integration and Solution Design | 16% | 60 |
| Configuration and Knowledge Management | 12% | 42 |
| Governance, Risk, and Responsible Use | 15% | 54 |
| Troubleshooting and Optimization | 10% | 36 |
| Total | 100% | 360 |
| Exam code | CCDV-F |
|---|---|
| Number of items | 53 |
| Item format | Multiple-choice and multiple-response; each item states how many responses to select |
| Time limit | 120 minutes |
| Passing score | Scaled score of 720 on a scale of 100-1,000 |
| Validity period | 12 months from the date the credential is awarded |
| Delivery | Proctored: online proctored and/or test center |
| Domains | 8 |
Official blueprint weighting beside the number of questions this site holds for each domain, so you can check the bank follows the blueprint rather than take it on trust.
| Domain | Official weight | Questions here |
|---|---|---|
| Agents and Workflows | 14.7% | 24 |
| Applications and Integration | 33.1% | 51 |
| Claude Code | 3.1% | 6 |
| Eval, Testing, and Debugging | 2.6% | 3 |
| Model Selection and Optimization | 16.8% | 27 |
| Prompt and Context Engineering | 11.0% | 18 |
| Security and Safety | 8.1% | 12 |
| Tools and MCPs | 10.6% | 18 |
| Total | 100% | 159 |
Yes, entirely. There is no paid tier, no ads, no upsell and nothing to buy. The material was written during exam preparation and costs very little to keep online.
Not for the practice pages on this site - those run entirely in your browser and store nothing. An account is only needed for the full question banks, because the quiz engine has to store your answers to score them and show you which domains you are weak in. That is the only reason.
All four: Claude Certified Associate - Foundations (360 questions), Claude Certified Developer - Foundations (159), Claude Certified Architect - Foundations (360), and Claude Certified Architect - Professional (189). That is 1,068 questions in total, all of them live now.
No, and that matters. Every question is written from scratch against Anthropic's official exam guides. None of them come from a real exam. They are not a shortcut past the exam; they test the same decisions the exam tests.
A solutions architect who has sat and passed these exams, with the credentials published on Credly and linked in the footer. Every bank is built the same way: from scratch, against the official exam guides, to the published blueprint weighting. More on the method is on the about page.
This is the part nobody mentions until you try. These exams are open to Anthropic partners only. Your employer has to be a member of the Anthropic Partner Network, and you register with your corporate email address at that company - there is no way to sign up as an individual, and a personal address will not get you registered.
The Associate and Architect Foundations tracks have a practice quiz per exam domain that pulls ten questions at random from the bank, plus timed practice sets and a full exam simulator on a 120-minute timer. The Developer and Architect Professional tracks currently have the timed practice sets and the full simulator, without the per-domain quizzes.
One full-length, timed run per track: 60 questions in 120 minutes for Associate and Architect Foundations, 53 questions in 120 minutes for Developer, and 63 questions in 120 minutes for Architect Professional. You get a score and a per-domain breakdown at the end.
Yes - and not just for the right answer. Every option has its own explanation of why it is or is not correct, plus an overall note on the reasoning. That is the part missing from nearly everything else in this niche, so it gets the most care here.
No. This is an independent community resource. Claude and Anthropic are trademarks of Anthropic, PBC. This site is not affiliated with, endorsed by, or sponsored by Anthropic, and it links to Anthropic's official exam guides rather than reproducing them.