Single-cell 3D genome mapping offers insights into chromatin architecture, yet the inherent sparsity and high dimensionality of contact maps hinder reliable identification of chromatin loops at the single-cell level. Here, we present CellLoop, a single-cell chromatin loop detection algorithm based on a density-based center detection framework that integrates intra-cellular and neighboring inter-cellular contacts via a re-voting strategy. Applied to Dip-C data from the mouse brain, CellLoop demonstrates superior accuracy based on the genome spatial distance and compartment signals at the single-cell level. It reveals cell-specific chromatin loops linked to transcriptional regulation and improved cell state classification. In HiRES embryogenesis data, CellLoop refines cell subtype definitions by minimizing cell cycle effects. Integrating GAGE-seq with MERFISH in mouse cortex, CellLoop redefines spatial domain functions through chromatin loop dynamics. Together, CellLoop enables robust, scalable, and biologically meaningful single-cell chromatin loop detection.