LangGraph is an open-source Agent framework developed by the LangChain team. It is powerful, especially after the v1.0 release, with comprehensive framework capabilities. v1.0 is the stable version promised by the development team, and the API is expected to remain stable, making it a great time to learn.
According to PyPI records,
LangGraph 1.0.0was released on October 18, 2025. It has only been a little over a month. Most code available today is based onv0.6, and this tutorial aims to bridge that gap.
1) Advantages of LangGraph
LangGraph is powerful. Memory, MCP, guardrails, state management, multi-agent — it has it all. Agent requirements are extremely complex, which can lead to framework code bloat and uncontrolled complexity. How does LangGraph 1.0 solve this problem? It introduces Middleware, a new feature. Middleware is a hook function. By embedding middleware into workflows, efficient extension and customization can be achieved.
2) Advantages of This Tutorial
You might ask, why read this when there’s the official documentation?
However, this tutorial is an interactive Jupyter Notebook-based tutorial, with each chapter independently runnable. The official documentation often only provides core code snippets that cannot be run directly. That is to say, the official documentation focuses on conceptual explanations, while this tutorial focuses more on practical implementation.
Promise: This tutorial is entirely based on
LangGraph v1.0, with no legacyv0.6code.
I. Table of Contents¶
Quick overview of this tutorial:
| No. | Chapter | Main Content |
|---|---|---|
| 1 | Quickstart | Create your first ReAct Agent |
| 2 | StateGraph | Create workflows using StateGraph |
| 3 | Middleware | Implement four features using custom middleware: budget control, message truncation, sensitive word filtering, PII detection |
| 4 | Human-in-the-loop | Implement human-in-the-loop using built-in HITL middleware |
| 5 | Memory | Create short-term memory and long-term memory |
| 6 | Context Engineering | Manage context using State, Store, and Runtime |
| 7 | MCP Server | Create MCP Server and integrate with LangGraph |
| 8 | Supervisor Pattern | Two methods to implement supervisor pattern: tool-calling, langgraph-supervisor |
| 9 | Parallel | How to implement concurrency: node parallelism, @task decorator, Map-reduce, Sub-graphs |
| 10 | RAG | Three ways to implement RAG: vector retrieval, keyword retrieval, hybrid retrieval |
| 11 | Web Search | Implement web search: DashScope, Tavily and DDGS |
| 12 | Deep Agents | Brief introduction to Deep Agents |
| 13 | Gradio APP | Develop streaming chat agent application based on Gradio |
| 14 | Appendix: Debug Page | Introduce the debug page provided by langgraph-cli |
II. Debug Page¶
langgraph-cli provides a debug page that can be launched quickly.
langgraph devSee details: Appendix
III. Further Reading¶
✨ If you find this project helpful, please Star this repository: luochang212