Claude Advanced Labs · Agents

Claude Advanced Labs - Agents

We publish technical resources for the AI security and development community. This advanced lab track restores the earlier autonomous-agent curriculum: a setup module, ten advanced modules, and one production-style orchestrator, and hands-on patterns for agent infrastructure, tool use, orchestration, reliability, and governance.

Start Module 0 → Quick Reference Cloud Architect Practice Exam
Companion Resource

Use NotebookLM alongside this lab

Claude API Development and Integration Guide

The source material for this lab lives in a NotebookLM notebook. Use it to ask follow-up questions, explore topics in more depth, or get a plain-English explanation of any concept while working through the modules.

Open in NotebookLM
Setup + 11 Modules

What you'll build

Each module builds on the last. By the end you'll have a working blueprint for an autonomous AI marketing agent and the agent patterns behind production Claude systems.

00
Dev Environment Setup
Install VS Code, set up Jupyter notebooks, and connect the Anthropic Python SDK to verify your environment before Module 1.
01
Foundations of Claude Orchestration
Master the Messages API stateless design, create a reusable Managed Agent, and control conversation flow with the agentic loop stop_reason pattern.
02
Tool Design & Schema Safety
Wire web search with Dynamic Filtering, enforce strict tool schemas, and understand the ZDR implications of code execution.
03
Advanced Agent Orchestration
Pair Sonnet (executor) with Opus (advisor) mid-generation for strategic guidance without extra round-trips.
04
Scaling with Message Batches
Run high-volume campaigns at a 50% discount with the Batches API, extended 300k-token outputs, and custom_id result tracking.
05
Claude Code & Repository Context
Master CLAUDE.md hierarchy, path-scoped rules with YAML frontmatter, reusable Agent Skills, and CI/CD pipeline integration.
06
MCP Integration
Connect external systems via the MCP Connector, expose resources as a content catalog, and manage credentials securely with .mcp.json env vars.
07
Prompt Engineering for Extraction
Guarantee JSON output with compiled grammars, prevent hallucination with nullable fields, and understand the citations incompatibility constraint.
08
Reliability & Error Handling
Use programmatic tool calling, structured error responses with isRetryable signals, and PostToolUse hooks to normalize pipeline outputs.
09
Compliance & Data Privacy
ZDR eligibility matrix for all features, HIPAA schema caching safeguards, and the 1.1x US-inference pricing multiplier.
10
Advanced Reliability & Compaction
Manage 1M-token contexts with server-side compaction and design resilient error recovery.
11
Final Project: Enterprise Orchestrator
Capstone that ties everything together: Managed Agents, MCP, hub-and-spoke subagents, advisor, batches, hooks, compaction, and case facts. Includes code starters and a full answer key.
✓
Cloud Architect Practice Exam
60 questions across all 5 domains. Randomized every attempt. Score breakdown and NotebookLM study prompt included.

Ready to build?

Open the NotebookLM guide in a second tab, then work through the advanced agent labs in order.

Begin Module 0 →