Callous-0923/agent-study

36章AI Agent全栈课程：从ReAct循环到Claude Code逆向、MCP/A2A协议、RAG、DSPy、生产可观测性——全部为可运行Python文件。

What it solves

Many developers want to learn AI agents but face scattered, unstructured resources. This course provides a from-zero, systematically designed path that covers both core theory and practical engineering, with a strong emphasis on job interview preparation and production readiness. It organizes 37 chapters into 7 progressive layers, and each chapter is a self-contained runnable Python file that doubles as complete lecture notes and executable code.

How it works

The course is structured in 37 chapters grouped into 7 layers: fundamentals, engineering practice, deep technical analysis, production & frontier, advanced architecture, core skill reinforcement, and expert-level advanced. You clone the repo, run a setup script, optionally configure an OpenAI-compatible API key (needed only for early chapters), and then run each chapter's Python file to see lecture content and demonstrations. Most later chapters (Ch8-12, Ch14-28) rely only on the Python standard library and require no API key. The full material is also available as a live website with responsive dark-blue design, so you can read and explore in the browser without installing anything.

Who it’s for

Targeted at fresh graduates, career-switching engineers, and any developer who wants to systematically master AI agents. It is especially useful for job seekers because it includes 20 high-frequency interview questions, marks interview key points per chapter, and offers multiple learning paths (from scratch, interview sprint, and product building).

Highlights

  • 37 chapters, 22000+ lines of code, 60+ runnable examples
  • Each chapter is a .py file that is both a complete lecture and runnable code
  • Most chapters beyond Ch8 require no API key and only use the Python standard library
  • Broad topic coverage: RAG deep dive, MCP protocol, A2A protocol, tool calling internals, model routing, semantic cache, agent security, DSPy, streaming architecture, observability, and more
  • Interview-oriented: marks high-frequency exam points and answer frameworks
  • 7 progressive layers from theory to expert-level practices
  • MIT license, free to use, modify, and distribute

Related

  • Project
    42-evey/hermes-plugins

    Hermes-plugins is a set of verified plugins and skills for Hermes Agent, adding autonomy, quality checks, memory, observability, and integrations. It’s designed to run with any model provider and comes with CI validation.

  • Project
    springvoiceswell/semrush-ai-tool

    An open-source MCP server and CLI that connects AI assistants to live Semrush data for automated SEO analysis, keyword research, and competitor intelligence.

  • Project
    Johell1NS/browser-search

    A self‑hosted skill that lets AI agents search with SearXNG, browse with a fast REST‑API Firefox (Camofox), and fall back to a stealth Chromium (CloakBrowser) to bypass anti‑bot challenges, all without external API keys.

  • Project
    ankimcp/anki-mcp-server

    Anki MCP Server is a Node‑JS bridge that implements the Model Context Protocol, exposing 53 JSON‑based tools (deck, card, note, media, model, and GUI operations) so AI assistants like Claude Desktop, Cursor, or cloud‑based ChatGPT can read, edit, and review Anki flashcards. It works via STDIO (for desktop AI clients) or HTTP, and offers a managed tunnel for remote access. Install via npm, an MCPB bundle, or from source, and configure the client to launch the server as a subprocess.