wtf is agentic-engineering-handbook?

This repository is a learning roadmap for people who want to build AI agent systems using tools from OpenAI, Anthropic (Claude), and Google. It collects 121 official resources, blog posts, videos, and code tutorials from those vendors and organizes them into a structured progression from beginner to production-level work. The roadmap is divided into phases. Phase 0 starts with the basics: building a simple agent loop from scratch in Python, showing how a model plus tools plus one repeating loop is the foundation of almost every AI agent. Phase 1 covers agent fundamentals like tool use and handoffs between agents. Phase 2 focuses on MCP (Model Context Protocol), a standard for connecting AI models to external data and tools. Phase 3 covers how agents handle memory and context, including how to give them persistent instructions via configuration files like CLAUDE.md or AGENTS.md. Phase 4 goes into longer-running agent setups and the infrastructure needed to run them reliably. Each phase includes a short reading list drawn from original vendor documentation, some supplementary material, and a practical build exercise. The exercises escalate in complexity, from a basic customer service routing bot to multi-agent evaluation systems. The code samples are Python scripts and include starter files, environment setup templates, and working examples tied to each tutorial step. The tutorials are markdown documents kept alongside the code in the same folder structure. This is a reference and study guide, not a runnable product or installable framework. Readers are expected to follow the linked external resources, work through the exercises on their own, and use this repo as an organized index and study planner. The full README is longer than what was shown.