Abstract: The future of AI systems lies in large-scale, heterogeneous agent collaboration to solve increasingly diverse and complex tasks. This collaborative paradigm promises greater flexibility and scalability but brings a fundamental challenge: how to balance collaboration efficiency with quality of tasks. Existing MCP and A2A architectures struggle to achieve this trade-off, facing either scalability bottlenecks or unstable coordination. To address these issues, we propose MoFedNet, a novel semantic link-guided model collaboration system. MoFedNet features a dual-layer architecture with centralized monitoring and decentralized collaboration. Its core SL mechanism enables scalable, efficient, and dynamic collaboration among heterogeneous models while supporting intelligent task orchestration and network evolution. We enhance MoFedNet with three key modules. Link-to-Link Semantic Protocol (L2L) organizes interactions over the hypergraph. Contextual Memory Enhanced Retrieval (CoMER) introduces structured memory modules to store and retrieve representations across abstraction levels. Evolutionary Local-Global Optimization Module (EvoLOM) drives continuous improvement by accumulating memory and optimizing the hypergraph. Simulation on over 15,000 nodes shows that MoFedNet attains 99.8 percent task success under lower resource and time costs compared with MCP and A2A, outperforms in the proposed QE index, and maintains stable temporal and spatial complexity.