A recent Chinese technical post proposes a novel architecture for the AI agent ecosystem: an Agent Factory platform and an A2A (Agent-to-Agent) network. The author argues that current agent products fall into three categories—monolithic agents, limited orchestration tools, and closed ecosystems—each with significant limitations. The proposed solution separates concerns cleanly: the Agent Factory serves as a development and runtime base for building agents, while the A2A network acts as an open marketplace where agents can discover and interact with each other. The two products communicate via a standardized A2A protocol and are independently deployable. This design addresses key challenges such as interoperability, scalability, and vendor lock-in. For developers and founders building agent infrastructure, this blueprint offers a practical, modular approach that could influence how future agent ecosystems are structured. The post includes detailed technical considerations for protocol design, deployment isolation, and market dynamics.
This post outlines a dual-product design: an Agent Factory platform for building and running agents, and an A2A network for agent-to-agent marketplaces. It identifies three categories of current agent products and their limitations, then proposes a clean separation via the A2A protocol. The design is technically detailed and commercially relevant for anyone building agent infrastructure.