github.com / WenyuChiou / awesome-agentic-ai-zh
The AI Agent learning map¶
From "what is an LLM and how are tokens counted" all the way to building your own multi-agent systems.
🤖 First: what is an AI Agent?¶
An AI Agent is an AI system that can decide what to do next and take action toward a person's goal. Once given a goal, it reads the current situation, uses tools when needed, then continues, corrects course, stops, or hands control back based on the result. It can do work automatically on a person's behalf, but only within clear rules and permissions.
Pick a learning track¶
-
Track A — CLI power user
Push CLI agents like Claude Code to their limit: workflows, productionization.
-
Track B — Agent builder
From tool calls all the way to multi-agent systems and your own MCP server.
Stage 0 through Stage 8, plus the Stage 7.5 reading stop¶
-
Stage 0 — Foundations
Check Python, Git, and API basics first; skip this stop if you are ready.
-
Stage 1 — LLM basics
Tokens, context, choosing a model.
-
Stage 2 — Prompt engineering
Say what you want so the model delivers reliably.
-
Stage 3 — Tool Use & Your First Agent Loop
Give the LLM tools; write your first agent.
-
Stage 4 — Workflow Graphs & Agent Frameworks
LangGraph, AutoGen, Agents SDK — which to pick.
-
Stage 5 — Claude Code ecosystem
CLI agents, MCP, Skills, subagents.
-
Stage 6 — Memory & RAG
Let an agent remember and retrieve.
-
Stage 7 — Agent Production Engineering
Make loops, workflow graphs, harnesses, and multi-agent work reliable.
-
Stage 7.5 — Advanced reading stop
Pick one advanced idea at a time and decide whether your system needs it.
-
Stage 8 — Agent interfaces
Computer Use, Browser Use, Sandbox.
Trilingual, with hands-on exercises in every stage. For the full intro and table of contents → project overview.