<p>Multi-agent collaborative mapping technology holds significant application value within the domain of simultaneous localization and mapping. Currently, achieving a comprehensive representation of complex scenes while efficiently ensuring global consensus among multiple agents remain a formidable challenge. In recent years, implicit scene representation, characterized by its lightweight structure, has been employed in mapping tasks. Nevertheless, existing methods are predominantly constrained to single-agent scenarios, and fail to adequately address specific issues such as geometric detail loss, reconstruction artifacts, and inadequate global collaboration support. Consequently, a novel multi-agent collaborative implicit mapping model, termed Topo-Map, is proposed, which is underpinned by a topological communication protocol. Specifically, Topo-Map integrates a hierarchical sampling strategy coupled with a surface-aware adaptive grid optimization method, enabling a screening mechanism for fuzzy grids and the establishment of prior constraints. The demands for hole filling and artifact suppression are optimally met by this dual strategy. Meanwhile, an adaptive enhanced feature aggregation module is incorporated, which is engineered to capture muti-scale features. Leveraging upon this foundation, the consensus alternating direction multiplier method is implemented within a neighborhood topological communication graph, constructed through a gaussian soft neighbor heuristic mechanism, thereby achieving information sharing and gradient synchronization. Furthermore, a parameter consistency loss is introduced to constrain weight disparities, ensuring global mapping consensus while preserving the local autonomy of individual agents. Experimental results demonstrate that Topo-Map achieves enhanced stability with notable robustness and performance gains in collaborative implicit mapping, offering a scalable solution for SLAM in intricate scenes.</p>