Claude Certified Architect: Foundations
Work in progress (WIP) — This playbook is a draft. The AI Service Anthropic certification path, timing, and Forge navigation are not finalized. Use Snowflake and Omni guidance in the certifications README as the active priorities until L&D publishes the final map.
| Cert name | Claude Certified Architect: Foundations |
| Issued by | Anthropic |
| Service line | AI Service |
| Q2 target date | April 21–28, 2026 (take this first — fastest Q2 win) |
| Cost | 99 otherwise |
| Duration | 120 minutes |
| Questions | 60 multiple-choice |
| Passing score | 720 / 1000 |
| Registration | Anthropic Academy on Skilljar |
| Prerequisites | 6+ months experience building with Claude APIs, Agent SDK, Claude Code, and MCP in production |
| Back to cert index | README.md |
Start here — this is the fastest Q2 win
The AI Service team has been building with Claude APIs, MCP servers, and Claude Code in production. The 6-month prerequisite is already met. This certification is the most natural fit for the team’s existing experience — take it first, in April, before the Snowflake prep ramps up.
Why this certification
Anthropic’s Claude Certified Architect: Foundations is the first official architecture-level technical certification for Claude. It validates the ability to make informed design decisions when building production-grade applications with Claude — agentic systems, MCP integrations, Claude Code workflows, and structured output pipelines.
Holding this cert demonstrates to clients and partners that Brainforge’s AI Service team does not just use Claude; they can architect with it. It also positions the team for the Claude Partner Network, which brings deal support and co-selling opportunities with Anthropic.
Step 1: Check access and cost
Before registering for the exam, check whether Brainforge qualifies for the free exam via the Claude Partner Network.
Partner Network access request: anthropic.skilljar.com/claude-certified-architect-foundations-access-request
If Brainforge is a Claude Partner Network member, the exam is 99 per attempt. Request access in Week 1 — do not wait until you are ready to take the exam.
Step 2: Complete the prep courses
All prep courses are free on Anthropic Academy. They are hosted on Skilljar — create an account (separate from your Anthropic account) to track progress and earn completion certificates.
Complete them in this order. Each course builds on the previous one.
Core prep sequence
| # | Course | Link | Time (est.) | Why it matters for the exam |
|---|---|---|---|---|
| 1 | Building with the Claude API | anthropic.com/claude-with-the-anthropic-api | ~3–4 hours | Comprehensive coverage of the API surface — messages API, tool use, streaming, system prompts; directly maps to Structured Output and Advanced Patterns domains |
| 2 | Introduction to Model Context Protocol | anthropic.com/introduction-to-model-context-protocol | ~2–3 hours | MCP server and client fundamentals — tools, resources, prompts; maps to the Tool Design & MCP domain (18%) |
| 3 | Model Context Protocol: Advanced Topics | anthropic.com/model-context-protocol-advanced-topics | ~2–3 hours | Sampling, notifications, file system access, transport mechanisms; completes the MCP domain |
| 4 | Claude Code in Action | anthropic.com/claude-code-in-action | ~2 hours | Claude Code workflow integration; maps directly to the Claude Code domain (20%) |
| 5 | Introduction to Agent Skills | anthropic.com/introduction-to-agent-skills | ~1.5 hours | Building, configuring, and sharing Skills in Claude Code; part of the Agentic Architecture domain |
| 6 | Introduction to Subagents | anthropic.com/introduction-to-subagents | ~1.5 hours | Subagent patterns in Claude Code, task delegation, context management; completes the Agentic Architecture domain |
| 7 | AI Fluency: Framework & Foundations | anthropic.com/ai-fluency-framework-foundations | ~1–2 hours | Context on responsible AI collaboration; lighter weight, good for rounding out before the exam |
Total estimated prep time: 13–16 hours across the 7 courses.
Optional additional courses
| Course | Link | When to take it |
|---|---|---|
| Claude 101 | anthropic.com/claude-101 | Only if you want a light refresher before starting the core sequence |
| Introduction to Claude Cowork | anthropic.com/introduction-to-claude-cowork | Useful context but not directly tested; skip if time-constrained |
Step 3: Study the exam domains
| Domain | Weight | What to study |
|---|---|---|
| Agentic Architecture & Orchestration | ~27% | Multi-agent coordination patterns, task routing, Agent SDK patterns, context window management in long-running agents, human-in-the-loop design, failure handling and retry logic |
| Claude Code | ~20% | Claude Code task loop, Skills (SKILL.md structure, triggers, context injection), subagent creation and lifecycle, memory and context strategies, responsible multi-step agent design |
| Tool Design & MCP | ~18% | MCP server architecture (tools vs resources vs prompts), tool schema design (JSON Schema), error handling in tool responses, client-server transport (stdio vs SSE), security considerations |
| Structured Output & Data Processing | unweighted (est. 15–20%) | JSON mode, tool_use for structured extraction, schema validation patterns, handling partial outputs, batch processing |
| Advanced Patterns & Production Optimization | unweighted (est. 15–20%) | Prompt caching, streaming with text_delta events, token counting and cost management, rate limit handling, retry with exponential backoff, production monitoring patterns |
What to go deep on for AI Service
The Agentic Architecture domain (~27%) is the most heavily weighted and maps exactly to what the AI Service team builds for clients. Know the patterns cold: how to coordinate multiple Claude agents on a shared task, how to route sub-tasks, and how to design for failure. This is not a recall question — the exam tests design judgment.
The Claude Code domain (20%) and MCP domain (18%) together make up nearly 40% of the exam. Both are areas where the Brainforge team has hands-on experience (Cursor skills = SKILL.md, Cursor agents = subagents, platform MCPs). Connect what you already do in your daily work to the exam concepts.
Step 4: Review the certification prep materials
Anthropic provides official practice questions and sample questions through exam prep resources.
Official certification page: claudecertifiedarchitect.net Preparation guide: claudecertifiedarchitect.net/how-to-prepare.html Free practice questions and study guide: claudecertifications.com
Use the practice questions after completing the course sequence — not before. They are most useful for identifying gaps, not for initial learning.
Step 5: Sit the exam
Register: Through Anthropic Academy on Skilljar — follow the access link from Step 1, or go to the official cert page once Partner Network access is confirmed.
Exam day checklist:
- Partner Network access confirmed (or $99 payment processed)
- All 7 core prep courses completed and marked done in Skilljar
- Practice questions reviewed; confidence on all 5 domains
- Government-issued photo ID ready (if required by proctoring)
- Quiet, uninterrupted space
- 120 minutes, 60 questions — 2 minutes per question average; do not rush
- Read each question carefully — agentic design questions are scenario-based and require choosing the best architectural approach, not just the correct fact
After passing
- Share your result in
#ai-learning— log against M1.2 (AI Service milestone). - Your completion certificate from Skilljar is available immediately. Your Anthropic credential/badge will follow per Anthropic’s issuance process.
- Begin SnowPro Core prep — see snowpro-core.md.
The combination of Claude Certified Architect + SnowPro Core + SnowPro Specialty: Gen AI is the full Q2 AI Service cert achievement.
What comes next
| After Q2 | Path |
|---|---|
| SnowPro Core + Gen AI | See snowpro-specialty-gen-ai.md |
| Deeper Claude work | Anthropic Academy continues to add courses — watch anthropic.com/learn/courses for new additions |
| External contribution | Subject Matter Expert program via Anthropic Academy once more certs are available |
Owner: Brylle Girang | Slack: #ai-learning
Last updated: 2026-04-01