Claude Certified Architect – Foundations (CCAR-F)

Learn how to design and coordinate Claude-powered agents that can plan tasks, invoke tools, process results, and continue working until a task is completed or requires intervention. Understand when to let the model decide the next step and when application code must enforce a workflow or safety constraint.

Key topics

• Agentic loops and termination conditions

• Multi-agent orchestration and task decomposition

• Subagent invocation and context passing

• Workflow enforcement, handoffs, and guardrails

• Agent SDK hooks

• Session state, persistence, and resumption

What this domain tests: Selecting the right orchestration pattern, deciding when to delegate work, handling incomplete tasks, and balancing model-driven decisions with deterministic application logic.

Understand how Claude interacts with external systems through tools, APIs, and Model Context Protocol (MCP). Effective tool design requires more than declaring a function: the model must understand what the tool does, provide valid inputs, interpret its output, and respond appropriately when execution fails.

Key topics

• Tool interfaces, descriptions, and input schemas

• Input validation and structured error responses

• Tool distribution and tool-selection strategies

• MCP server and client integration

• Built-in tools and their appropriate use

What this domain tests: Designing reliable tool contracts, choosing between integration approaches, handling tool errors, and deciding how external capabilities should be exposed to Claude.

Explore how to configure Claude Code for repeatable, maintainable development workflows. Learn how project instructions, reusable skills, scoped rules, execution planning, and automation influence Claude's behavior across different tasks and codebases.

Key topics

• CLAUDE.md hierarchy, scope, and modular organization

• Custom commands and skills

• Path-specific rules and conditional instructions

• Plan Mode versus direct execution

• Iterative refinement and review workflows

• CI/CD integration

What this domain tests: Choosing the appropriate configuration mechanism, applying instructions at the correct scope, selecting a suitable execution workflow, and integrating Claude Code into software development processes.

Learn how to guide Claude toward accurate, consistent, and machine-readable responses. This domain covers the design of clear instructions, the use of examples, structured response patterns, and validation strategies that help applications handle model outputs reliably.

Key topics

• System prompts with explicit criteria

• Few-shot prompting and examples

• Structured output and tool-based response patterns

• Output validation, retries, and feedback loops

• Batch processing strategies

• Multi-instance processing and multi-pass review

What this domain tests: Selecting an appropriate prompting technique, enforcing output requirements, validating responses, and designing recovery or review strategies when the first result is insufficient.

Understand how to preserve relevant information and maintain reliable behavior as tasks grow more complex. This includes managing limited context, resolving ambiguity, preventing failures from spreading across agents, and ensuring that generated conclusions are supported by trustworthy information.

Key topics

• Context-window management

• Escalation and ambiguity resolution

• Error propagation in multi-agent systems

• Codebase exploration and context degradation

• Human review and confidence calibration

• Information provenance and multi-source synthesis

What this domain tests: Choosing what information to retain, determining when to seek clarification or human intervention, diagnosing cascading failures, and maintaining traceability and reliability in AI-generated results.