Callous-0923/agent-study

This open-source project is a self-contained, 36-chapter course on building production AI agents, structured entirely as runnable Python files rather than docs or slides. The scope is unusually honest about what production actually means: it goes past the "build a chatbot" demo layer and into observability, the A2A and MCP protocols, DSPy prompt optimization, and a reverse-engineering study of Claude Code internals. That last piece alone is worth pulling the repo — most agent curricula skip the part where you learn how real systems are structured by reading one. The interview-orientation is a pragmatic framing: each chapter maps to something a hiring team might actually probe on, which doubles as a forcing function for the author to make the material concrete rather than theoretical. Reservation: the README is Chinese-first, which adds friction for English-only readers, and the repo looks like a solo learning project rather than a maintained course. Treat it as a reading list with runnable examples, not a structured MOOC. -> Best for: AI engineer or solo founder who wants to go from "I have seen agent demos" to "I understand the production stack"