The self-labeling protein HaloTag, reacting specifically and irreversibly with bio-orthogonal substrates, has become a popular tool for protein labeling. Particularly when combined with a variety of tailor-made fluorescent substrates, HaloTag excels in high and super-resolution microscopy, where fluorescent proteins reach their limits. Through genetic fusion, HaloTag enables protein tracking with high spatial and temporal resolution in living cells. However, direct fusion of HaloTag can interfere with their expression, folding, trafficking or function. In addition, tagging endogenous proteins with large tags using CRISPR/Cas9 knock-in approach remains relatively laborious. Therefore, a simple and efficient way to bypass direct tagging with the full-length HaloTag would be of great value. A high-affinity split HaloTag system, incorporating one of the two split fragments as a small peptide, could achieve that. A split HaloTag system was previously developed in the group, which contains a large inactive protein fragment (i.e. cpHaloDelta1) whose activity can be complemented by a short peptide (Hpep1). However, the millimolar affinity between these two fragments does not allow spontaneous self-complementation. Therefore, I engineered cpHaloDelta and Hpep to improve their intrinsic affinity using yeast surface display screening and subsequently validated the improved variants in mammalian cell culture and fluorescence microscopy analysis. The split pair (cpHaloDelta3/Hpep5) showed nanomolar affinity and demonstrated spontaneous self-complementation ability in vitro and in cells. This pair resulted in the best performance for endogenous target visualization among all the tested variants. Hpep5 is a 14-mer peptide, enabling straightforward endogenous protein tagging in cells through the CRISPR/Cas9 knock-in approach. This allowed the endogenous tagging of seven targets, which are expected to have varying expression levels, in U-2 OS cells overexpressing cpHaloDelta3. The proper localization of these targets was confirmed through confocal fluorescence microscopy with high contrast. Live-cell stimulated emission depletion microscopy (STED) imaging was performed on three of these targets, where the performance of split HaloTag was comparable to that of HaloTag. Furthermore, the split HaloTag system is compatible with expansion microscopy (ExM), allowing access to super-resolution imaging when live-cell imaging is not required and STED equipment is not available. Finally, certain variants of the identified Hpeps influenced the brightness of the fluorophore on the reconstituted split HaloTag surface. This property was exploited to develop a multiplex imaging approach using fluorescence lifetime imaging (FLIM). Overall, I anticipate that the high-affinity split HaloTag system will have multiple applications, especially in, but not limited to, fluorescence live-cell imaging.