Claude Certified Architect - Foundations (CCAR-F)¶

Anthropic's first official Claude certification, launched March 12, 2026. It validates the foundational knowledge needed to design and build production Claude systems: agentic architecture, Claude Code configuration, prompt engineering, tool design, Model Context Protocol (MCP) integration, and context and reliability patterns.

It targets solution architects, AI engineers, and developers who have hands-on experience building with the Claude API, Claude Code, and MCP. Anthropic positions it as the entry point of the Architect track; Claude Certified Architect - Professional sits above it.

Exam Overview¶

Registration goes through the 📖 Anthropic Partner Academy - membership in the (free) 📖 Claude Partner Network is required, and a personal email address is blocked at checkout. Exams are scheduled with 📖 Pearson VUE; badges are issued via Credly. Retakes: 14-day wait after attempt 1, 30 days after attempt 2, 90 days after attempt 3, maximum 4 attempts in a rolling 12 months.

Exam Domains¶

Domain Breakdown¶

1 - Agentic Architecture (27%)¶

The highest-weighted domain. You should understand how to design agentic systems with Claude, including when to use agents vs simple prompts, multi-agent orchestration, and production reliability patterns.

Key Concepts: - Agentic design patterns (tool use loops, multi-step reasoning, orchestration) - Agent architectures (single agent, multi-agent, supervisor/worker patterns) - Agentic workflows (plan-execute-reflect loops) - Error handling and recovery in agentic systems - Decision criteria for agents vs simple prompt chains - Production considerations (cost management, latency, reliability) - Claude's extended thinking in agentic contexts

2 - Claude Code Configuration (20%)¶

Claude Code is Anthropic's official CLI and IDE integration for Claude. This domain covers configuring, customizing, and managing Claude Code for individual and team development workflows.

Key Concepts: - Claude Code CLI setup, installation, and configuration - CLAUDE.md file hierarchy (project-level, user-level, enterprise-level) - Settings and permissions model - Hooks system (pre-tool and post-tool execution hooks) - Custom slash commands - IDE integrations (VS Code extension, JetBrains plugin) - MCP server configuration within Claude Code - Team workflow configuration

3 - Prompt Engineering & Structured Output (20%)¶

The art and science of crafting effective prompts for Claude, plus extracting structured data from responses.

Key Concepts: - Prompt engineering best practices specific to Claude - System prompts vs user prompts and their roles - Chain-of-thought and reasoning techniques - Structured output extraction (JSON mode, tool use for structured data) - XML tags for prompt organization and clarity - Few-shot and many-shot prompting techniques - Prompt caching for cost optimization - Multimodal prompting (vision, PDF processing)

4 - Tool Design & MCP Integration (18%)¶

The Model Context Protocol (MCP) is a core part of Claude's extensibility story. This domain covers designing tools, building MCP servers, and integrating external capabilities into Claude workflows.

Key Concepts: - Model Context Protocol (MCP) architecture and fundamentals - MCP components (clients, servers, transports - stdio, SSE, streamable HTTP) - Building MCP servers (tools, resources, prompts) - Tool design best practices (naming conventions, descriptions, JSON schemas) - Claude API tool use (function calling) - Tool choice modes (auto, any, specific tool forcing) - Parallel tool use - Error handling in tool calls

5 - Context & Reliability (15%)¶

Managing Claude's context window effectively, implementing reliability patterns, and building production-grade applications.

Key Concepts: - Context window management - Long context best practices and strategies - Prompt caching for performance and cost - Extended thinking (Claude's reasoning mode) - Reliability patterns (retries, fallbacks, validation loops) - Token counting and budget management - Streaming and real-time response handling - Rate limiting, error handling, and API best practices

Study Approach¶

Recommended Path¶

1. Start with the official prep courses - The Anthropic Partner Academy hosts free preparation courses and the official exam guide. Complete those first; they define the exam scope.

2. Read the documentation - The official docs at docs.anthropic.com are the primary source of truth. Focus on the API reference, guides, and cookbooks.

3. Build hands-on projects - This is applied material. You need practical experience building with the Claude API, Claude Code, and MCP to internalize it.

4. Practice with scenarios - Work through the scenario-based questions in this guide. The goal is applied judgment, not memorization.

5. Review and fill gaps - Use the fact sheet and notes in this guide to identify and fill knowledge gaps.

Time Investment¶

• 4 weeks is the recommended study period for someone with existing Claude experience
• 6-8 weeks for those newer to Claude but with general AI/ML background
• 2-3 hours per day of focused study is ideal

Key Study Resources¶

Study Materials in This Guide¶

Practice questions: resources/practice-questions/anthropic-claude-architect-foundations.md

Exam Tips¶

1. Read every word - Scenario questions have important details buried in the context. Skim at your peril.
2. Eliminate first - Most questions have 1-2 obviously wrong answers. Eliminate those, then reason through the remaining options.
3. Think production - Favor production-ready, reliable, cost-effective solutions over clever hacks.
4. Watch for "most appropriate" - Many questions ask for the BEST answer, not just a correct one. Multiple options may work, but one is better.
5. Time management - 120 minutes for 60 questions is 2 minutes each. Flag and move on.
6. Agentic Architecture is king - At 27% weighting, prioritize it.

The Anthropic Certification Track¶

CCAR-F is one of four official Claude certifications:

A typical architect path: CCAR-F first, then CCAR-P once you have production experience. Developers shipping Claude apps usually sit CCDV-F instead. The Prompt Engineering Specialist guide in this repo remains a self-directed deep-dive that supports all four exams.