Hello-Agents: The Complete Guide to the "Building Agents from Scratch" Tutorial

Hello-Agents is a comprehensive 16-chapter tutorial on building AI agents from scratch — from LLM foundations to multi-agent systems, with hands-on projects. By Datawhale China, the largest open-source AI education community in China. Theory + practice, covering agent paradigms, memory, context engineering, communication protocols, and reinforcement learning. 26,300+ stars, Python.

Hello-Agents on GitHub

---

What Is Hello-Agents?

2025 is the "Agent Year" — the focus has shifted from training bigger models to building smarter agent applications. Yet systematic, practice-oriented tutorials are scarce. Hello-Agents fills this gap with a 16-chapter curriculum that takes you from zero to building multi-agent systems.

The tutorial focuses on AI-Native Agents (truly AI-driven) rather than workflow-driven agents (Dify, Coze, n8n). You'll penetrate framework surfaces, understand core architectures, learn classic paradigms, and build your own multi-agent applications.

• Stars: 26,300+ ⭐
• Forks: 2,949
• Contributors: 58
• Language: Python
• Author: Datawhale China
• Website: datawhalechina.github.io/hello-agents
• Topics: agent, llm, rag, tutorial
• Format: Online reading + PDF download

---

16-Chapter Curriculum

Part 1: Foundations (Ch 1-3)

Chapter	Title	Content
1	Introduction to Agents	What are agents, core concepts
2	Agent History	Evolution from rule-based to LLM-powered
3	LLM Foundations	Large language models as agent backbone

Part 2: Architecture & Building (Ch 4-7)

Chapter	Title	Content
4	Classic Agent Paradigms	ReAct, Chain-of-Thought, Plan-and-Execute
5	Low-Code Agent Building	Dify, Coze, n8n platforms
6	Framework Development	Using existing frameworks
7	Build Your Own Framework	Create your agent framework from scratch

Part 3: Advanced Topics (Ch 8-12)

Chapter	Title	Content
8	Memory & Retrieval	RAG, long-term memory, vector search
9	Context Engineering	Managing context windows effectively
10	Agent Communication Protocols	MCP, inter-agent messaging
11	Agentic-RL	Reinforcement learning for agents
12	Performance Evaluation	Benchmarking and testing agents

Part 4: Projects (Ch 13-16)

Chapter	Title	Content
13	Smart Travel Assistant	End-to-end agent project
14	Automated Deep Research Agent	Research automation
15	Building a Cyber Town	Multi-agent simulation
16	Graduation Project	Final capstone

---

Hello-Agents vs Alternatives

Category: This is a comprehensive AI agent building tutorial/course.

Feature	Hello-Agents	LangChain Docs	OpenAI Cookbook	AutoGen Tutorial
Focus	Full agent curriculum	Framework docs	API recipes	Multi-agent framework
Stars	26.3K ⭐	~100K ⭐	~60K ⭐	~40K ⭐
Chapters	16	N/A	Varied	N/A
Theory + Practice	✅ Both	Practice only	Practice only	Practice only
Agent History	✅	❌	❌	❌
Build Own Framework	✅ Ch 7	❌	❌	❌
Context Engineering	✅ Ch 9	Partial	❌	❌
Agentic-RL	✅ Ch 11	❌	❌	❌
Capstone Projects	✅ 4 projects	❌	❌	Examples
PDF Download	✅	❌	❌	❌
Community Blog	✅	❌	❌	❌

When to choose Hello-Agents: You want a structured, university-style curriculum for learning agent development from scratch — theory + practice + capstone projects.

When to choose LangChain Docs: You want framework-specific documentation for building with LangChain.

When to choose OpenAI Cookbook: You want practical API recipes for specific OpenAI use cases.

---

Conclusion

Hello-Agents is the most comprehensive open-source agent-building curriculum available — 16 chapters covering everything from LLM foundations to multi-agent simulations, with a deliberate focus on AI-native agents rather than workflow automation. At 26.3K stars, it's the go-to resource for anyone who wants to transform from an LLM "user" into an agent "builder."

Explore Hello-Agents on GitHub

---

Resources

• GitHub Repository
• Online Reading
• Datawhale Community