agent_learning
A systematic AI Agent development tutorial covering LLM agents, RAG, tool use, memory systems, multi-agent systems, LangChain, LangGraph, MCP, and agentic RL.|
What is agent_learning?
A systematic AI Agent development tutorial covering LLM agents, RAG, tool use, memory systems, multi-agent systems, LangChain, LangGraph, MCP, and agentic RL.|从零开始学 AI Agent 开发 | 系统、全面、实战导向的 Agent 开发教程 | 每日自动追踪 arXiv 最新论文 | Learn AI Agent Development from Scratch
Overview of agent_learning
What it is
Agent_learning is an open-source AI Agent textbook and learning repository. It provides 23 chapters across foundations, core capabilities, frameworks, multi-agent systems, production, and capstone projects. The content is maintained in both English and Chinese, with 188 Markdown pages per language, 330+ original SVG diagrams, and five interactive demos. The repository also includes a runnable reference-agent implementation with tools, memory, security gates, evaluation, an MCP server, a FastAPI service, and 16 tests.
Key features
- 23 chapters covering topics from LLM fundamentals to production deployment, including RAG, tool use, memory systems, agentic RL, multi-agent systems, LangChain, LangGraph, and MCP.
- Bilingual content with 188 pages per language and 330+ original SVG diagrams.
- Runnable reference-agent baseline with a minimal ReAct loop, tool registry, offline FakeProvider, optional OpenAI provider, memory, prompt-injection guardrails, fail-closed permission checks, MCP server, FastAPI endpoints, streaming, evaluation harness, Dockerfile, and 16 tests that run without an API key.
- Paper-to-practice explanations covering ReAct, Reflexion, MemGPT/Letta, GraphRAG, GRPO, MCP, A2A, and more.
- The project follows five principles: mechanisms before frameworks; visuals must teach; research must lead to engineering insight; production claims must be honest; bilingual content stays aligned.
Use cases
- Learning AI Agent development systematically from first principles to production engineering.
- Teaching or referencing Agent architectures and implementation patterns.
- Prototyping a minimal Agent baseline with offline testing capabilities.
Getting started
See the repository README for full instructions. To run the reference-agent tests locally:bashcd reference-agentpython -m venv .venvsource .venv/bin/activatepip install -e .[dev]pytest -q
License and pricing
Released under the MIT License. Open source, no pricing information stated.
Based on the project's public README and repository metadata.