Show HN: AgentNexus – coordinate LLM agents by service boundary, not role (github.com)

A new project called AgentNexus has emerged, introducing a coordination architecture for large language model (LLM) agents that focuses on service boundaries rather than agent roles. This novel approach challenges existing multi-agent frameworks, such as ChatDev and MetaGPT, which are centered around role definitions within a simulated organizational structure. By organizing around service granularity, AgentNexus mirrors real-world software system architecture, allowing each service to register as a sub-project, manage versioned documents, and receive notifications for document changes, thus streamlining code modification without human intervention. The significance of AgentNexus lies in its potential to enhance collaboration among LLM agents in software development. It features versioned document storage with SHA-256 deduplication, a robust publish-subscribe notification system, and diff-aware updates that enable targeted code changes based on document modifications. Additionally, it supports an AI-driven onboarding process for integrating agents seamlessly into projects. With tools such as full-text search and a browser-based dashboard for exploring projects, AgentNexus offers a comprehensive framework for developing sophisticated AI-driven software solutions, poised to facilitate more efficient and organized workflows within the AI/ML community.