Claude Certified Architect – Foundation (CCAR-F)
The architectural design credential for solution architects, lead engineers, and AI technical consultants building production-grade enterprise Claude solutions.
| Exam Code | Cost | Duration | Questions | Passing Score | Delivery Format |
|---|---|---|---|---|---|
| CCAR-F | $125 USD | 120 Minutes | 60 Questions | 720 / 1000 (72%) | Pearson VUE (Online / Center) |
Important Firsthand Exam Insight: Scenario-Based Blocks
The CCAR-F exam uses a unique structure: 6 comprehensive real-world enterprise scenarios are established in the official syllabus, and 4 of those 6 are randomly selected on exam day. Each selected scenario has 15 sequential questions asked about it.
Strategy Tip: Read the specific question at the end of each item first before re-reading the lengthy scenario background to save time.
Who This Certification Is For
Solution architects, lead software engineers, and technical consultants responsible for sizing, designing, optimizing, and securing production LLM architectures using Anthropic models.
Core Exam Domains & Objectives
Click any domain title below to access full study material, official syllabus sub-sections, and 3 interactive scenario practice questions:
-
Domain 1: Prompt Caching Architecture & Cost ROI →
Mechanics of ephemeral caching (`cache_control`), minimum prefix token thresholds (1,024 for Sonnet/Opus; 2,048 for Haiku), 5-minute rolling TTL refresh, and cache write vs cache read economics. -
Domain 2: Latency Optimization, TTFT & Message Batches API →
Model selection matrices (Haiku vs Sonnet vs Opus), Token-to-First-Token (TTFT) reduction, output token pruning, and using the Message Batches API for 50% discount on async workloads. -
Domain 3: Multi-Agent & Orchestration Design Patterns →
Designing Router-Worker patterns, Evaluator-Optimizer loops, parallel sub-agent workers, and context window hygiene. -
Domain 4: Long-Context Window Management & Retrieval Architecture →
Sizing context windows (200k tokens), mitigating "lost-in-the-middle" recall issues, chunking strategies, and combining RAG with prompt caching. -
Domain 5: Production Reliability, Security Guardrails & Circuit Breakers →
Enforcing deterministic safety guardrails in execution code, least-privilege tool access, rate-limit resilience, and circuit breaker patterns.
Key Architectural Mechanics to Memorize
1. Prompt Caching Economics & Rules
- Cache Write: 25% surcharge on initial token write.
- Cache Read: 90% discount on subsequent requests hitting the cached prefix.
- TTL: 5-minute rolling window (extended every time the cache is read).
- Ordering: Static context (system prompts, large tools, reference docs) must be at the beginning; dynamic user query must be at the end.
2. Message Batches API
Designed for non-realtime, asynchronous workloads (e.g. nightly backfills, evaluations). Delivers responses within 24 hours at an automatic 50% cost discount across all Claude models.
Self-Assessment & Mock Exams
Self-Assessment Quiz
Take our 10-question architectural self-assessment covering prompt caching, latency SLAs, multi-agent patterns, and reliability.
Quiz →Full-Length Mock Exams
Practice with 6 full mock tests (360 scenario questions) modeled specifically on the 6 official CCAR-F enterprise scenario tracks on Udemy.
Mock Tests →Official Anthropic Resources for CCAR-F
Anthropic provides architect learning tracks and detailed documentation on prompt caching and batch APIs. Access official resources below: