Agent Design Patterns Society
A design language for production AI agent systems.
ADPS is a small but serious research community sedimenting the field experience of Chinese engineers — agent practice, software engineering methodology, open-source harness work, enterprise deployment feedback — together with the recurring architectural problems and messy experience of the agent era, into a pattern language the world's engineers can use, debate, review, and pass on.
The 28-pattern matrix · 7 cognitive functions × 6 execution topologies
A note on Choreography. Choreography is sometimes listed as a topology. We treat it instead as a coordination mode that operates across topologies — closer in spirit to the Teams pattern than to an independent execution shape — so it is described in the patterns text rather than given its own column.
What we do
ADPS organizes recurring agent-system architecture problems into a coherent pattern language. The work is structured around three deliverables:
- The pattern catalog — 28 named patterns on a 7×6 matrix (cognitive function × execution topology), each with a runnable implementation and an engineering slice from a real production codebase.
- The pattern selection methodology — a 6-step process plus 5 laws that turn "I'm building an agent for X" into a candidate pattern list in under fifteen minutes.
- The production harness — open-source reference implementations that show how patterns compose under cost, latency, and governance constraints.
The framework
The two-axis framework — cognitive function × execution topology — was introduced in A Two-Dimensional Framework for AI Agent Design Patterns (Huang & Zhou, arXiv:2605.13850) and developed further in the Manning book Designing AI Agents. ADPS now stewards the catalog as a community asset.
Founders
Four founding members, four forces: architecture language, software engineering, open-source ecosystem, enterprise deployment.
-
Jia Huang — Framework · Writing · Global Narrative
A*STAR Singapore. Author, Designing AI Agents (Manning). -
Bingsheng Ru (茹炳晟) — Software Engineering · Agent System · R&D Productivity
Industry Research Consultant, Tencent Research Institute. Vice Chair, CCF TF Multi-Agent Systems Engineering SIG. -
Willem Jiang (姜宁) — Open Source · Harness · Workflow
Apache Software Foundation contributor. DeerFlow. -
Bo Liang (梁博) — Enterprise SaaS · FinTech · Production Feedback
VP, Shanghai Dongfang Yiteng. Banking & FinTech SaaS, agent transformation.
Advisor
- Joey Tianyi Zhou (周天异) — A*STAR IAIC Agent Chapter Director. Associate Professor at NUS. Co-author of the two-axis framework paper.
Expert Group
- Haofen Wang (王昊奋) — Tenured Professor, Tongji University. Founding rotating chair of OpenKG. Knowledge graphs & knowledge-enhanced LLMs (OpenSPG / KAG-Thinker / MemOS); 12,700+ citations.
- Haili Zhang (张海立) — Author of the LangGraph / LangChain practical guides. LangChain Ambassador. LFAPAC open-source evangelist.
Assets
- arXiv:2605.13850 — the position paper introducing the two-axis framework.
- github.com/huangjia2019/agent-design-patterns — runnable catalog of 28 patterns. Will migrate under the ADPS organization.
- github.com/huangjia2019/designing-ai-agents — companion code to Designing AI Agents (Manning).
-
Manning · Designing AI Agents
— the design-pattern catalogue for production AI agents.
Reader codes at Manning:designing-ai(eBook / MEAP, mainland China only — $20.99) ·MLHuang4(50% off all formats, valid Jun 9–23).
Reach out
ADPS is open to engineers and organizations with field experience building production-grade agent systems — at scale, in regulated industries, or under hard reliability constraints. Membership is by contribution, not application.
Fill the three fields below and we'll get in touch through WeChat. Nickname is fine — no real name required.
Prefer GitHub? Open a discussion or issue on agent-design-patterns.