I started coding in 2016. No mentor, no Jira tickets, no one to tell me how to decompose a requirement. I was handed a module, end to end: upload frontend, server, S3, database, echo. I drew the diagram and I wrote every line. One brain, one pair of hands, one chain from start to finish.

A decade later, the chain is MCP and A2A. The planning loop is ReAct. The documentation lookup is RAG. I no longer drive just my own hands — I drive a squad of specialist Agents, orchestrated into a self-sufficient workforce.

What I build is not a chat interface. A chat dies when the conversation ends. I build a living system: one that senses, decides, and acts. Draw the boundary, it runs inside it.

This is not automation (that's a 20-year-old word). It is not replacement (that's a fear-driven narrative). It is amplification. I call the shots; they fight the war.

The line I care about isn't between one tool and another. It's between things that forget and things that grow. Between a Coding Agent — a stateless expert you start from scratch with every session — and a Living Agent, one that remembers, learns, and compounds experience across time. The longer you use it, the better it gets. Not because you configured it, but because it grew with you.

That's the pattern I build toward: systems where time is a moat, not a reset button.

Focus      AI full-stack engineering · Multi-agent orchestration
Direction  AI applications · Workflow orchestration · Platform engineering
Credo      Human-amplifying AI, not human-replacing AI
Languages  TypeScript · Python
Based in   Chengdu, China

• Production AI applications that ship, not prototypes that stall
• Multi-agent systems that plan, execute, and self-correct without hand-holding
• Content generation and campaign delivery pipelines at scale
• Internal platforms with measurable gains in efficiency, stability, and cost

• Languages: TypeScript, Python
• AI Agent: MCP, A2A, ReAct, LangChain, Dify, LLMOps
• Backend: Node.js, FastAPI, Prisma, BullMQ, PostgreSQL, Redis
• Frontend: React, Next.js, Taro, Flutter, React Native, Electron
• Infra: Docker, DevOps, Playwright (CDP automation)
• Engineering: Monorepo governance, observability, automation

• Multi-agent system (MAS) architecture design
• ReAct planning and execution agent development
• MCP protocol — unified LLM tool-calling standard
• A2A protocol — inter-agent communication and distributed collaboration
• Sandbox isolation — secure code execution in Docker containers
• Playwright browser automation — CDP protocol control
• Context engineering — memory persistence and session state management

• GitHub: zhangshichuan
• Email: zsc.guru@qq.com
• LinkedIn: maxzhang1010

方向        AI 全栈工程 · 多智能体系统指挥
聚焦        AI 应用 · 工作流编排 · 平台工程
信条        放大人的 AI，不是替代人的 AI
主要语言    TypeScript · Python
所在地      中国成都