LangChain & LangGraph: A Systematic Course

Understand Agent orchestration from first principles: State Graph + Nodes + Conditional Edges.

What is this / 这个项目是什么

An open-source learning project for developers who can program but have not yet worked with LLMs / Agents. The goal is to help you truly understand how Agents work, so you can build complex applications on a solid foundation — not just copy templates.

Core Thesis / 核心价值主张

We believe most mainstream Agent frameworks (CrewAI, AutoGen, etc.) can be mapped onto LangGraph's model of StateGraph + Node + Conditional Edge.

Master LangChain and LangGraph, and you master the underlying paradigm of Agent orchestration.

That is also what sets this project apart from "yet another LangChain tutorial": instead of listing APIs, we help you build a general mental model so you can understand and design any Agent system.

Learning Path (13 Phases) / 学习路线

Phase	Topic	Key Output
1	LangChain Core & Environment (skippable primer)	First Chain
2	Chains, Memory & RAG Basics	Multi-step chains, RAG
3	Agent Principles: ReAct to Loops	Agent loop
4	Tools & Function Calling	Tool use
5	Multi-Agent Patterns & Comparison	Framework → graph mapping
6	LangGraph Orchestration Core	StateGraph in practice
7	State, Memory & Human-in-the-loop	Interruptible workflows
8	Production Deployment	Serving + observability
9	Capstone Project & Full Review	End-to-end system
10	Independent Memory Layer: episodic + semantic + traceability	Traceable memory
11	MCP Tool Interface: standard-protocol tools	agent ↔ tool
12	Multi-Agent Collaboration (A2A): supervisor + worker	agent ↔ agent (in-process)
13	A2A Standard Protocol: Agent Card + JSON-RPC	agent ↔ agent (cross-process)

Progressive design: P1–P9 build intuition with concrete frameworks and converge on the "everything is a graph" paradigm; P10–P13 are advanced hands-on chapters — memory layer → MCP → native multi-agent → A2A standard protocol — completing the agent-interconnection picture of a production multi-agent system.

Getting Started / 如何开始

1. Set up Python & dependencies (see Phase 1 · Python/Env Primer).
2. Configure .env to choose a model (OpenAI / DeepSeek / Qwen / local Ollama).
3. Run your first Chain.

Bilingual / 双语

This docs site is available in Simplified Chinese (default) and English (you are here). Switch via the language toggle at the top.