Autogenesis Protocol Introduces Self-Evolving AI Agent Framework
The Autogenesis Protocol (AGP) presents a novel structure for self-evolving AI agent systems, addressing the shortcomings of A2A and MCP. It consists of two distinct layers: the Resource Substrate Protocol Layer (RSPL), which treats prompts, agents, tools, environments, and memory as versioned resources, and the Self Evolution Protocol Layer (SEPL), offering a closed-loop interface for suggesting and implementing enhancements with auditing features. The Autogenesis System (AGS) employs a multi-agent architecture to oversee evolving components. As detailed in arXiv preprint 2604.15034v2, AGP tackles issues related to lifecycle management and version tracking, with the goal of minimizing fragile code and improving structured evolution interfaces while ensuring audit trails and fallback mechanisms.
Key facts
- Autogenesis Protocol (AGP) is a self-evolution protocol for AI agent systems
- Addresses limitations in existing protocols like A2A and MCP
- Decouples what evolves from how evolution occurs
- Resource Substrate Protocol Layer (RSPL) models prompts, agents, tools, environments, and memory as versioned resources
- Self Evolution Protocol Layer (SEPL) provides closed-loop operator interface for improvements
- Autogenesis System (AGS) is a self-evolving multi-agent system built on AGP
- Research documented in arXiv preprint 2604.15034v2
- Aims to reduce brittle glue code and monolithic compositions in agent systems
Entities
Institutions
- arXiv