Machine Learning

Claude Certified Architect - Foundations (CCA-F)

Full preparation for the CCA-F certification exam - agentic architecture, tool design and MCP, Claude Code, prompt engineering and structured output, and context management, mapped objective by objective to the v1.0 blueprint.

Advanced2466176 lessons10 free lessonscca-fcertificationclaudeagentic-architecturemcp

Claude Certified Architect - Foundations (CCA-F)

Complete preparation for the CCA-F exam: 60 scenario-based questions in 120 minutes, scored 100-1000, with 720 to pass.

The exam does not test whether you can recite an API. It tests whether you can read a scenario and choose an architecture - and nearly every question turns on one lens: where model discretion belongs, and where you need deterministic enforcement instead. That lens runs through all five domains, and this course teaches it explicitly rather than leaving you to infer it.

Twenty modules cover the blueprint end to end:

• Agentic architecture (D1, 27%) - the action loop, stop_reason, loop control and runaway protection, task decomposition, multi-agent designs, and when a deterministic workflow simply beats an agent
• Tool design and MCP (D2, 18%) - tool contracts and descriptions, MCP foundations, and what changes when MCP goes to production
• Claude Code (D3, 20%) - the instruction hierarchy, commands, skills and subagents, hooks, and the Agent SDK
• Prompt engineering and structured output (D4, 20%) - core and advanced prompting, JSON and schema-constrained output, validation loops, and evaluation
• Context management and reliability (D5, 15%) - context window economics, escalation and error propagation, and long-running work with confidence routing

Then Scenario Labs for applied practice, and Exam Readiness with three full-length timed mocks, per-objective remediation, and exam-day tactics.

Content verified against live documentation as at 27 August 2026, against blueprint v1.0 (effective July 2026).

Ten topics across six modules are free to preview, deliberately sampling every domain rather than only the opening modules. The full course is a one-time purchase with lifetime access.

What you will learn

• Read a scenario and decide where model discretion belongs and where enforcement must be deterministic
• Design the agent loop properly - stop_reason handling, termination conditions, turn caps and circuit breakers
• Judge when a deterministic workflow beats an agent, and defend the choice from stated constraints
• Write tool contracts and MCP servers that survive production, not just a demo
• Use Claude Code deliberately: instruction hierarchy, commands, skills, subagents and hooks
• Get reliable structured output, with validation gates instead of hope
• Budget the context window and design escalation and error propagation that degrade safely
• Recognise the four trap patterns the exam uses, and pace a 60-question scenario paper

Prerequisites

• Comfortable calling an LLM API from code - the exam assumes you have built something, not just read about it
• Basic familiarity with JSON, HTTP and one general-purpose programming language
• No prior certification required

Course outline

Module 1: Orientation and Exam Mechanics

Orientation and Exam Mechanics - 6 topics.

• What CCAR-F Certifies, Who It Is For, and the Prerequisite Bar 11 min read
• Exam Format, Scaled Scoring, Per-Domain Reporting, and What 720 Means 12 min read
• The Scenario Pool: How One Narrative Anchors a Block of Questions 12 min read
• The Central Lens - Model Discretion Versus Deterministic Enforcement 13 min read
• The Four Trap Patterns 13 min read Pro lesson
• Weighted Study Planning and the Per-Objective Readiness Bar 12 min read Pro lesson

Module 2: Foundations of Agentic Systems

Foundations of Agentic Systems - 11 topics.

• What Makes a System Agentic: Autonomy, Tool Access, and the Action Loop 11 min read
• Agents, Workflows, and Conversational Systems - The Real Differences 12 min read
• Tool Use as the Mechanism of Agency 12 min read Pro lesson
• The Loop: Perceive, Reason, Act, Observe 12 min read Pro lesson
• How stop_reason Drives the Loop - tool_use Versus end_turn 13 min read Pro lesson
• Loop Control: Termination Conditions, Iteration Limits, and Escalation Exits 13 min read Pro lesson
• Runaway Loops: Turn Caps, Budget Ceilings, and Circuit Breakers 13 min read Pro lesson
• When a Deterministic Workflow Beats an Agent 13 min read Pro lesson
• Choosing an Architecture From Stated Constraints 13 min read Pro lesson
• Model Selection as an Architectural Decision: Capability, Latency, and Cost Routing 14 min read Pro lesson
• Anti-Pattern Review: Agentic Designs That Should Have Been Workflows 14 min read Pro lesson

Module 3: Task Decomposition and Planning

Task Decomposition and Planning - 8 topics.

• Turning a Complex Goal Into Executable Subtasks 13 min read Pro lesson
• Decomposition Shapes: Hierarchical, Sequential, and Parallel 13 min read Pro lesson
• Sequential Pipelines - Design, Coupling, and Failure Behaviour 13 min read Pro lesson
• Parallel Execution: Fan-Out, Fan-In, and Synchronisation 13 min read Pro lesson
• Replanning Mid-Execution When the World Changes 13 min read Pro lesson
• Ambiguous and Incomplete Specifications: Clarify, Assume, or Escalate 14 min read Pro lesson
• Verifying a Plan Before Executing It 13 min read Pro lesson
• Decomposition Depth: When Subtasking Stops Paying for Itself 13 min read Pro lesson

Module 4: Multi-Agent Architecture

Multi-Agent Architecture - 11 topics.

• The Orchestrator: Directing, Delegating, and Aggregating 13 min read Pro lesson
• Subagent Design: Scope, Authority, and Isolation Boundaries 13 min read Pro lesson
• Topologies: Hub-and-Spoke, Pipeline, and Peer-to-Peer 13 min read Pro lesson
• Selecting a Topology From the Problem Shape 13 min read Pro lesson
• Designing Agent-to-Agent Message Schemas 13 min read Pro lesson
• Handoff Protocols and Continuity Across Agent Boundaries 13 min read Pro lesson
• Structured Handoff Packages: What the Receiving Agent Needs 13 min read Pro lesson
• In-Context State Versus External Memory 13 min read Pro lesson
• Session Continuity and Resumption Across Turns and Failures 13 min read Pro lesson
• Multi-Agent Cost, Latency, and Fit 13 min read Pro lesson
• When a Single Agent Is the Right Answer 13 min read Pro lesson

Module 5: Reliability, Enforcement, and Escalation

Reliability, Enforcement, and Escalation - 12 topics.

• Three Failure Classes: Tool, Reasoning, and Environment 13 min read Pro lesson
• Detecting Failure Before It Cascades 13 min read Pro lesson
• Retry Policy: What Is Retryable, What Is Not, and Idempotency 13 min read Pro lesson
• Fallback Chains and Graceful Degradation 13 min read Pro lesson
• Why Prompt Instructions Are Insufficient on High-Stakes Paths 13 min read Pro lesson
• Building Programmatic Validation and Enforcement Layers 13 min read Pro lesson
• Orchestration-Layer Safeguards Versus Model-Side Instruction 13 min read Pro lesson
• Escalation Triggers: Valid Patterns and Invalid Ones 13 min read Pro lesson
• Interruption Points and Human Review Workflows 13 min read Pro lesson
• The Terminal-State Contract: Every Request Resolves or Escalates 13 min read Pro lesson
• PreToolUse and PostToolUse Hooks as Enforcement Points 14 min read Pro lesson
• Observability for Agents: Logging, Tracing, and Telemetry 13 min read Pro lesson

Module 6: Tool Design Fundamentals

Tool Design Fundamentals - 12 topics.

• How Claude Selects a Tool: Descriptions as the Routing Surface 12 min read
• Writing Tool Descriptions That Route Correctly 12 min read Pro lesson
• Input Schema Design for Precision and Reliability 12 min read Pro lesson
• Required Versus Optional Parameters, Enums, and Constraints 12 min read Pro lesson
• Strict Tool Use and Input Validity 12 min read Pro lesson
• Structured Tool Errors: Returning Failure the Model Can Act On 12 min read Pro lesson
• Tool-Level Retry and Partial-Success Semantics 12 min read Pro lesson
• Scoping Tools Per Agent: Why Fewer Is Better 12 min read Pro lesson
• Selection Heuristics and Common Selection Failures 12 min read Pro lesson
• Forcing and Constraining Selection With tool_choice 12 min read Pro lesson
• Tool Search for Large Catalogs 12 min read Pro lesson
• Overlapping and Colliding Tools: Disambiguating a Crowded Catalog 12 min read Pro lesson

Module 7: MCP Foundations

MCP Foundations - 7 topics.

• What MCP Is and the Problem It Solves 12 min read
• The Three Primitives: Tools, Resources, and Prompts 12 min read Pro lesson
• Building a First MCP Server in Python 13 min read Pro lesson
• Exposing Tools Through MCP 13 min read Pro lesson
• Exposing Resources 12 min read Pro lesson
• MCP Prompt Templates and Their Use Cases 12 min read Pro lesson
• Implementing Clients and Connecting to Servers 13 min read Pro lesson

Module 8: MCP in Production

MCP in Production - 12 topics.

• STDIO Transport: Local Servers and Process Communication 12 min read Pro lesson
• Streamable HTTP Transport: Remote Servers and Deployment 12 min read Pro lesson
• Sampling: Servers Making Model Calls Through the Client 12 min read Pro lesson
• Notifications, Roots, and Progress Reporting 12 min read Pro lesson
• Boundaries, Access Scoping, and Least Privilege 12 min read Pro lesson
• MCP Server Scopes: Local, Project, and User Configuration 13 min read Pro lesson
• Authentication Patterns for MCP Servers 12 min read Pro lesson
• Versioning and Compatibility Across Server Changes 12 min read Pro lesson
• Reliability: Timeouts, Health Checks, and Degraded Servers 12 min read Pro lesson
• Trust Boundaries: Prompt Injection Through Tool Results and Untrusted Servers 14 min read Pro lesson
• The Built-In Toolset - What Ships With Claude 12 min read Pro lesson
• Built-In Versus Custom: The Boundary Decision 12 min read Pro lesson

Module 9: Claude Code Fundamentals

Claude Code Fundamentals - 7 topics.

• What Claude Code Is and How the Session Model Works 12 min read
• The Tool System: Reads, Edits, Bash, and Search 12 min read Pro lesson
• Onboarding Claude Code to an Existing Codebase 12 min read Pro lesson
• Adding and Managing Context Deliberately 12 min read Pro lesson
• Permission Modes and What Each One Permits 14 min read Pro lesson
• Surfaces: Terminal, IDE, Desktop, and Non-Interactive 12 min read Pro lesson
• Choosing a Surface for a Team's Workflow 12 min read Pro lesson

Module 10: Configuration and Instruction Hierarchy

Configuration and Instruction Hierarchy - 9 topics.

• The Memory Hierarchy: Enterprise Policy, Project, User, and the Deprecated Local File 13 min read Pro lesson
• Precedence and Conflict Resolution Across Levels 13 min read Pro lesson
• Nested Subtree Discovery and Scoped Instructions 12 min read Pro lesson
• @path Imports and Composing Instruction Files 12 min read Pro lesson
• The settings.json Hierarchy - User, Project, Local, and Enterprise Managed Policy 13 min read Pro lesson
• Permission Rules: Allow, Ask, and Deny 13 min read Pro lesson
• Excluding Files and Restricting Access to Sensitive Paths 12 min read Pro lesson
• Instruction Design: What Actually Gets Read and Followed 13 min read Pro lesson
• Configuration Patterns for Real Teams 13 min read Pro lesson

Module 11: Commands, Skills, and Subagents

Commands, Skills, and Subagents - 9 topics.

• Creating and Structuring Custom Slash Commands 13 min read Pro lesson
• Command Design Patterns and Distributing Them to a Team 12 min read Pro lesson
• What Skills Are and When a Skill Is the Right Container 13 min read Pro lesson
• Authoring Skills 12 min read Pro lesson
• Creating and Configuring Subagents 13 min read Pro lesson
• Designing Subagents for Delegation 12 min read Pro lesson
• The Mechanism Decision Matrix - Instructions, Commands, Skills, Subagents, Hooks, and Permissions 14 min read Pro lesson
• Plan Mode: When Planning First Pays 12 min read Pro lesson
• Iterating on a Plan: Approve, Edit, Replan 12 min read Pro lesson

Module 12: Hooks, Automation, and the Agent SDK

Hooks, Automation, and the Agent SDK - 8 topics.

• The Hook Lifecycle and the Event Surface 14 min read Pro lesson
• The Common Hook Events and What Each Intercepts 14 min read Pro lesson
• Implementing Hooks: Matchers, Shell Scripts, Exit Codes, Timeouts, and Gotchas 15 min read Pro lesson
• Hooks in Practice: Linting, Logging, and Safety Nets 14 min read Pro lesson
• Hooks as Deterministic Enforcement 14 min read Pro lesson
• The Claude Agent SDK and Session Control 14 min read Pro lesson
• Non-Interactive Mode and Claude Code in CI/CD 14 min read Pro lesson
• Least Privilege in Automated Contexts 14 min read Pro lesson

Module 13: Core Prompt Engineering

Core Prompt Engineering - 7 topics.

• Clarity, Specificity, and the Cost of Ambiguity 15 min read
• Instruction Decomposition and Constraint Design 15 min read Pro lesson
• The System Prompt: Purpose, Structure, and What Belongs in It 15 min read Pro lesson
• Role Assignment, Persona, and Context Setting 15 min read Pro lesson
• XML Tags for Structuring Complex Prompts 15 min read Pro lesson
• Document Structure Patterns for Multi-Part Tasks 15 min read Pro lesson
• Prompt Caching and Cache-Aware Prompt Layout 16 min read Pro lesson

Module 14: Advanced Prompting

Advanced Prompting - 8 topics.

• Why Examples Work: Pattern Matching and Format Anchoring 15 min read Pro lesson
• Designing High-Quality Few-Shot Examples 15 min read Pro lesson
• Chain-of-Thought and Visible Reasoning 15 min read Pro lesson
• Extended Thinking: When It Helps and When It Costs 16 min read Pro lesson
• Steering Output: Prefill, Stop Sequences, and Format Anchors 15 min read Pro lesson
• Temperature, top_p, and top_k in Practice 15 min read Pro lesson
• Designing Multi-Turn Interactions 15 min read Pro lesson
• Conversation State, Context Accumulation, and Turn Management 15 min read Pro lesson

Module 15: Structured Output and Evaluation

Structured Output and Evaluation - 12 topics.

• The Reliability Ladder for Structured Output 15 min read Pro lesson
• JSON Mode and Schema-Constrained Output 15 min read Pro lesson
• Schema Design for Reliable Extraction 15 min read Pro lesson
• Limits of Structured Output and Where It Breaks 15 min read Pro lesson
• Using Tool Definitions to Extract Structured Data 15 min read Pro lesson
• Nullable Fields and Not-Found Semantics: Schemas That Refuse to Fabricate 16 min read Pro lesson
• Sync Versus Message Batches: Choosing the Request Mode 15 min read Pro lesson
• Validation Loops and Hallucination Prevention 15 min read Pro lesson
• Validation Gates and Retry Strategy 15 min read Pro lesson
• Multi-Pass Review: When One Pass Is Not Enough 15 min read Pro lesson
• Building a Prompt Evaluation Workflow 15 min read Pro lesson
• Test Dataset Generation and Model-Based Grading 15 min read Pro lesson

Module 16: Context Window Economics

Context Window Economics - 9 topics.

• How the Context Window Works: Tokens, Limits, and Trade-offs 15 min read Pro lesson
• Context Economics: Cost, Latency, and Quality 15 min read Pro lesson
• The Stateless Messages API and What It Forces on Your Design 15 min read Pro lesson
• Why Summaries Lose Facts 15 min read Pro lesson
• Designing a Persistent Fact Block 15 min read Pro lesson
• The Memory Tool for Persistence 15 min read Pro lesson
• Tool-Output Bloat and Its Compounding Cost 15 min read Pro lesson
• Trim Strategies: Filter, Summarise, Reference 15 min read Pro lesson
• Caching as a Context Strategy, Not Just a Cost Strategy 15 min read Pro lesson

Module 17: Escalation and Error Propagation

Escalation and Error Propagation - 7 topics.

• Escalation Triggers: Designing the Boundary 15 min read Pro lesson
• The Handoff Payload: What the Next Actor Needs to Act 15 min read Pro lesson
• Where to Catch an Error: Locally or Up the Stack 15 min read Pro lesson
• Designing the Propagation Contract 15 min read Pro lesson
• Ambiguous Empty Results and the Failures They Hide 15 min read Pro lesson
• Disambiguation Patterns and Tool-Wrapper Discipline 15 min read Pro lesson
• Escalating to a Human: The Interface Contract 15 min read Pro lesson

Module 18: Long-Running Work and Confidence Routing

Long-Running Work and Confidence Routing - 9 topics.

• What Long-Running Work Actually Demands 15 min read Pro lesson
• Where Confidence Comes From and What It Means 15 min read Pro lesson
• Calibration: Making Confidence Correspond to Accuracy 15 min read Pro lesson
• Confidence Routing: Thresholds and Consequence 15 min read Pro lesson
• The Scratchpad: Durable State Outside the Context 15 min read Pro lesson
• Compaction Strategy for Long Sessions 15 min read Pro lesson
• Resumption and Side-Effect Safety 15 min read Pro lesson
• The Review Queue: Design and Sustainability 15 min read Pro lesson
• Putting It Together: A Long-Running System 15 min read Pro lesson

Module 19: Scenario Labs

Scenario Labs - 6 topics.

• Scenario Lab 1 - Customer Support Resolution Agent 16 min read Pro lesson
• Scenario Lab 2 - Multi-Agent Research System 16 min read Pro lesson
• Scenario Lab 3 - Claude Code Across a Development Team 16 min read Pro lesson
• Scenario Lab 4 - Structured Data Extraction Pipeline 16 min read Pro lesson
• Scenario Lab 5 - Enterprise MCP Integration 16 min read Pro lesson
• Scenario Lab 6 - Long-Running Analysis Workflow 16 min read Pro lesson

Module 20: Exam Readiness

Exam Readiness - 6 topics.

• Diagnostic Assessment and Interpreting Your Baseline 15 min read Pro lesson
• Domain Drills and the Per-Objective Remediation Loop 15 min read Pro lesson
• Timed Mock 1 - 20-Question Form 40 min read Pro lesson
• Timed Mock 2 - 20-Question Form 40 min read Pro lesson
• Timed Mock 3 - 20-Question Form 40 min read Pro lesson
• Exam-Day Tactics: Pacing, Elimination, and the Most-Correct Convention 15 min read Pro lesson

Browse all courses