Neuromorphic computing, with its emulation of the brain's efficient and low-energy information processing characteristics, has emerged as a potent solution to the energy efficiency and processes speed bottlenecks inherent in traditional computing architectures. In the face of issues such as the singularity of performance, fabrication difficulties, and weak integration found in heterogeneous structures of conventional neuromorphic devices, the novel twodimensional atomic-molecular heterostructures (2DAMH) offer new opportunities for the neuromorphic device domain. These heterostructures achieve stability and tunable functionality through the covalent modification of functional molecules on the surface of two-dimensional materials. This review summarizes the latest advances in 2DAMH with respect to electronic properties, synthesis strategies, and applications in neuromorphic devices, particularly highlighting the potential in customizing interfacial properties and simulating biological synaptic functions. Despite challenges in precision, scalability, and theoretical maturity associated with covalent modification, innovative research continues to emerge, actively seeking solutions and demonstrating immense potential for achieving more efficient and energy-saving computational models in neuromorphic intelligent systems.