Hi everyone,
I’ve been experimenting with Cursor’s rules system to build something beyond coding conventions — a sort of operating system for AI-involved software engineering, which I call co-agenticOS.
The idea is that we don’t just write rules for formatting or file structure, but for how humans and AI agents collaborate: planning, documenting, reviewing, and reflecting together.
What is co-agenticOS?
A structured collection of Abstract Rules, Instance Rules, and Templates that define:
-
Culture Layer: How engineers + agents communicate and maintain shared context
-
Workflow Layer: Plan → Implement → Summarize → Reflect loop
-
AI Routing Layer: Choosing the right model for each task (Reasoner / Builder / Fast Agent / Context Agent)
-
Documentation Layer: Plans, Sum-Logs, and Reflections as first-class outputs
Goal
To make Cursor a place where teams can share and evolve agentic-AI workflows — rules that capture not just what to code but how to collaborate with AI.
Why post here
-
I’m curious whether others are also designing rule-driven development cultures inside Cursor projects.
-
Would it make sense for the community to have a new category like “Workflow & Rules” to exchange rule systems and templates?
-
If you’ve built similar patterns (e.g., reflection logs, AI review loops, model switching rules), please share your experience!
Reference
Repository draft: github.com/keven1894/co-agenticOS (coming soon)
Concept note: “Culture is the New Compiler” – manifesto.md
Questions to the community:
-
Do you use Cursor’s
.rules.mdto define team culture or just coding rules? -
Would a shared rule registry (or template exchange hub) be useful?
-
What features would help Cursor support agentic development better (e.g., rule grouping, context priority, auto summaries)?
Thanks for reading! Excited to hear how others are evolving their AI workflows in Cursor 
— Keven Guan (@keven1894)