Exam Guide
Summarised from the official Claude Certified Architect – Foundations Exam Guide v1.0. If anything here disagrees with the official guide, the guide wins.
Format
- Exam code
- CCAR-F
- Items
- 60 — multiple-choice and multiple-response (each item states how many to select)
- Structure
- 4 scenarios drawn at random from a bank of 6
- Time limit
- 120 minutes
- Delivery
- Proctored by Pearson VUE — online or at a test center
- Passing score
- Scaled score of 720 on a 100–1,000 scale
- Results
- Pass/fail with scaled score, plus percent-correct by domain
- Fee
- $125 USD per attempt
- Validity
- 12 months from the date the credential is awarded
Domains
D1 · Agentic Architecture & Orchestration27%
D2 · Tool Design & MCP Integration18%
D3 · Claude Code Configuration & Workflows20%
D4 · Prompt Engineering & Structured Output20%
D5 · Context Management & Reliability15%
Scoring
- Scores are scaled and equated across exam forms; the 720 cut score comes from a standard-setting study. It is not a simple percentage.
- The score report shows pass/fail, the scaled score and percent-correct by domain.
- Guessing: the guide does not mention a penalty for wrong answers. “No penalty for guessing” is unverified — check the official Exam Policy.
Decision tables
Hooks vs prompt instructions
| Situation | Use |
|---|---|
| A business rule must hold every time (e.g. verify identity before a refund) | Programmatic enforcement: PreToolUse hook or code-level prerequisite |
| Block a tool call before it runs | PreToolUse hook (exit code 2 blocks) |
| Normalize, log or lint after a tool call | PostToolUse hook (cannot block) |
| Soft guidance where occasional deviation is acceptable | Prompt instructions |
Source: Guide Tasks 1.4–1.5; Claude Code hooks docs
Plan mode vs direct execution
| Task | Mode |
|---|---|
| Large-scale, multi-file or architectural change | Plan mode first |
| Several valid approaches; needs exploration before committing | Plan mode |
| Simple, well-scoped change (e.g. a single-file bug fix) | Direct execution |
| Plan agreed | Switch to direct execution to implement it |
Source: Guide Task 3.4
When to escalate to a human
| Signal | Action |
|---|---|
| Customer explicitly asks for a human | Escalate immediately |
| Policy gap or exception the agent is not authorized to decide | Escalate |
| No meaningful progress after reasonable attempts | Escalate with a structured case summary |
| Low self-reported confidence or negative sentiment alone | Not a reliable trigger — these are poorly calibrated proxies |
Source: Guide Task 5.2
Batch vs synchronous API
| Need | Choose |
|---|---|
| Blocking pre-merge check, developer waiting | Synchronous Messages API |
| Large overnight or non-urgent workload | Message Batches API (50% cost, up to 24 h, no latency SLA) |
| Match results to requests | custom_id per request |
| Multi-turn tool calling inside one request | Not supported in a batch — use synchronous calls |
Source: Guide Task 4.5; Message Batches docs
Exam-day checklist
- ✓Read the official Exam Guide, the Certification Terms and the Exam Policy
- ✓Name on your government ID matches your registration exactly
- ✓Run the Pearson VUE system test before an online exam
- ✓Clear workspace: no notes, phones, smart watches, headphones or second monitors
- ✓Stay in webcam view for the whole online session
- ✓Reschedule or cancel at least 24 hours ahead, or the fee is forfeited