63,802 stars for a Chinese tutorial that builds agents from scratch

What it does Hello-Agents is a comprehensive open-source tutorial by the Datawhale community that walks developers through building intelligent agent systems from the ground up. Spanning sixteen chapters, it starts with LLM fundamentals and agent history, then moves into implementing classic paradigms like ReAct, Plan-and-Solve, and Reflection, and culminates in multi-agent projects such as a travel assistant and a cyber-town simulation. The curriculum also covers advanced topics including memory systems, RAG, context engineering, agent protocols like MCP and A2A, and Agentic RL training from SFT to GRPO.

The interesting bit The project makes a deliberate philosophical distinction between workflow-driven “software engineering agents” (e.g., Dify, Coze, n8n) and genuinely AI-native agents, staking its entire approach on the latter. It even ships watermarked PDFs to preemptively sabotage marketing accounts that might otherwise resell the free material to beginners.

Key highlights

• A custom HelloAgents framework built from scratch using raw OpenAI API calls, not existing abstraction layers
• Practical coverage of emerging agent communication protocols (MCP, A2A, ANP) and systematic evaluation frameworks
• Community extras including interview question banks, GUI agent tutorials, web agent anti-scraping guides, and self-evolution loops
• Real-world capstone projects: a DeepResearch clone, an MCP-powered travel planner, and a simulated society (“Cyber Town”)
• Fully free with online reading and PDF downloads; the README explicitly states there is no paywall

Caveats

• The entire curriculum is in Chinese, which limits direct accessibility for non-readers
• While every chapter is marked complete (✅), the provided README is truncated, so granular quality and maintenance status of individual sections is unclear without inspecting the files directly

Verdict Worth bookmarking if you read Chinese and want to graduate from prompting LLMs to architecting agent systems. If you are looking for a quick, English-language reference or a no-code drag-and-drop solution, this is not your stop.