Claude Certified Architect (CCA-001) — AI Agent Mastery
Everyone says they 'know AI.' CCA-001 proves it. Every mission drops you into a live AWS Bedrock sandbox where you build, deploy, and validate real AI agent infrastructure. Master agentic loops, tool design, MCP servers, prompt engineering, and reliability patterns.
Learning Paths
Architect's Atoll
Agentic loops, tool sequencing, multi-agent fleets, decision routing, and Bedrock Guardrails. 5 advanced missions.
Tool Master's Reef
Tool description quality, structured errors, MCP servers, and dynamic registries. 4 advanced missions.
Navigator's Compass
CLAUDE.md hierarchy, slash commands, plan-execute pipelines, and CI/CD integration. 4 advanced missions.
Prompt Cartographer's Isle
Explicit criteria, few-shot patterns, structured output forcing, schema validation, and prompt chains. 5 advanced missions.
Lighthouse of Reliability
Token budgeting, human escalation, error propagation, and observability dashboards. 4 advanced missions.
All Labs
Build the Core Agentic Loop with Bedrock
Implement the agentic loop pattern with stopReason-driven iteration.
Multi-Step Tool Sequencing for AI Agents
Chain multiple tool calls with reasoning steps in sequence.
Build a Multi-Agent Coordinator System
Design coordinator-subagent fleets for complex AI task delegation.
Agent Decision Loops with Retry & Fallback
Implement decision routing, retry logic, and escalation in AI agents.
Bedrock Guardrails with Content Filters
Configure Bedrock Guardrails for safe, compliant AI applications.
Tool Description Quality for AI Agents
Why tool descriptions matter more than parameters for agent accuracy.
Structured Error Responses for AI Tools
Build tools with isError and isRetryable structured error signals.
Build an MCP Server with JSON-RPC
Create a Model Context Protocol server with JSON-RPC transport.
Dynamic Tool Registry with DynamoDB
Build a runtime tool registry that loads tools dynamically from DynamoDB.
CLAUDE.md Hierarchy & Rule Configuration
Configure Claude Code with hierarchical CLAUDE.md files and glob-based rules.
Build Custom Slash Commands for Claude Code
Create custom slash commands backed by a DynamoDB registry.
Plan-Then-Execute Pipeline with Approvals
Build a plan-then-execute workflow with human approval gates.
CI/CD with Headless Claude Code
Run Claude Code in CI/CD pipelines with the -p flag for non-interactive mode.
Bedrock Managed Prompts with Explicit Criteria
Write prompts with measurable criteria using Bedrock Prompt Management.
Build a Few-Shot Classifier & Extractor
Use few-shot prompting to classify text and extract structured data.
Force Structured Output via Tool Use
Use tool_use forcing to guarantee JSON-structured AI responses.
Schema Validation with Retry Loop
Validate AI outputs against JSON schemas and auto-retry failures.
Multi-Pass Review Pipeline (Generate-Review-Refine)
Chain generate, review, and refine prompts into a quality pipeline.
Context Manager with Token Budgeting
Build a DynamoDB-backed context manager with sliding window truncation.
Human Escalation for Low-Confidence AI
Build confidence scoring and SNS-based human escalation patterns.
Multi-Step Error Propagation Pipeline
Build error propagation that doesn't silently swallow failures.
AI Operations Dashboard with CloudWatch
Build latency, error, and cost dashboards for AI agent monitoring.
Ready to start the CCA-001 track?
Every lab runs in a safe AWS sandbox with automated validation. Complete labs show up in your verified portfolio.
Launch Cloud Edventures →