Claude Certified Architect – Foundations — Study Plan

Synthesis of the Anthropic EAP study guide the founder shared 2026-05-27. RDCO's own study-plan work product. Founder clarified 2026-06-05: keeping/studying the source guide for personal study is fine (no copyright concern — the line is distribution or productizing it, not having a copy to study). If the founder re-shares the raw EAP guide, it can be pulled into the vault as a study reference.

✅ Cert identity — confirmed 2026-05-27

Founder confirmed via iMessage 2026-05-27: "Claude Certified Architect – Foundations" IS the cert tied to the +$5k escalator (deadline 2026-11-22). The escalator memory's old "Partner Architect" label was wrong and has been corrected. No test time reserved yet (founder's action — likely needs partner-portal access first).

What it tests

A solution architect who can make sound tradeoff decisions building production Claude solutions. Core surface: Claude Code, Claude Agent SDK, Claude API, MCP. Candidate profile: 6+ months hands-on across those four, including production limitations. (No hard prerequisite certs named.)

Format & scoring

Multiple-choice, scenario-based: 4 scenarios drawn at random from a pool of 6; each scenario frames a question set.
Each question = 1 correct + 3 distractors → pick the single best answer.
No guessing penalty — answer every question (unanswered = wrong).
Scaled score 100–1,000; passing = 720, pass/fail against an SME-set standard.
Duration / # questions / cost / registration / EAP-window: not in this guide — the registration comms came (or will come) separately. Action: locate the EAP enrollment comms (portal login, codes, seats, exam window).

Exam domains (weighted)

#	Domain	Weight	Scope (paraphrased)
D1	Agentic Architecture & Orchestration	27%	agentic loops, coordinator-subagent patterns, subagent context passing, multi-step handoffs, SDK hooks, task decomposition
D3	Claude Code Configuration & Workflows	20%	CLAUDE.md, Agent Skills, slash commands, plan vs direct execution, MCP server integration
D4	Prompt Engineering & Structured Output	20%	JSON schemas, few-shot, extraction patterns, reliable structured output
D2	Tool Design & MCP Integration	18%	designing MCP tool/resource interfaces for backend integration
D5	Context Management & Reliability	15%	long-doc/multi-turn context, error handling, human-in-the-loop, self-evaluation

RDCO advantage: D2+D3 (38% combined) map directly onto Ben's own ~/.claude harness (CLAUDE.md, 26 skills, slash commands, 10 MCP servers) — near-free points. D1 (heaviest at 27%) is the study priority.

Study plan (now → 2026-11-22, part-time alongside phData)

~25 weeks. The day job IS the lab — Ben builds with Claude Code/SDK/MCP daily, so prioritize the gaps between "uses daily" and "can answer scenario tradeoff questions."

Phase 1 (Jun, wks 1–4): D1 Agentic Architecture (27%). Heaviest weight. Agentic loop (stop_reason tool_use vs end_turn), coordinator-subagent, Task tool, hooks (PostToolUse), enforcement-vs-prompt-guidance. Checkpoint: explain all D1 patterns from memory.
Phase 2 (Jul, wks 5–8): D3 Claude Code (20%) + D2 MCP (18%). Direct overlap with Ben's harness — fast wins. Checkpoint: design an MCP tool interface + a CLAUDE.md from scratch.
Phase 3 (Aug, wks 9–12): D4 Prompt Eng/Structured Output (20%) + D5 Context/Reliability (15%). JSON schemas, few-shot, HITL, self-eval. Checkpoint: full-domain self-quiz.
Phase 4 (Sep, wks 13–16): Scenario drills. Work all 6 scenario types end-to-end; map each to primary domains. Checkpoint: cold-read any scenario, name the patterns.
Phase 5 (Oct, wks 17–20): Mock exams + gap-fill. Target a comfortable margin above 720. Re-drill weakest domain.
Phase 6 (early Nov, wks 21–24): Final review; sit exam ~Nov 7–14. Buffer before the 11-22 deadline for one retake if needed. Do NOT schedule the exam in deadline week.

Open actions

Confirm with phData: does "Foundations" = the escalator cert (+$5k)?
Locate the EAP registration comms (window, cost, seats, portal).
Decide study cadence + whether Ray runs spaced-repetition quiz drills per domain.

[[project_phdata_cert_escalator_path]] — Snowflake GenAI Specialty (90d) → this Anthropic cert (180d, +$5k)
[[2026-05-26-skillopt-self-evolving-agent-skills]] — D1/D3 agentic-architecture material RDCO already runs
[[2026-05-18-agentway-harness-engineering-claude-code-design-guide]] — D3 Claude Code config depth

claude certified architect foundations study plan