Protein kinases play a crucial role in many cellular signaling processes, making them one of the most important families of drug targets. But selectivity put a barrier at the design of kinase inhibitors. Fragment-based drug design strategies have been successfully applied to develop novel selective kinase inhibitors. However, the complicate kinase-fragment interaction and fragment-to-lead process pose challenges to fragment-based kinase discovery. Here, we developed a web source KinaFrag to investigate kinase-fragment interaction space and perform fragment-to-lead optimization. KinaFrag contained 31464 fragments from reported kinase inhibitors, which involved 3244 crystal fragment structures and 7783 crystal kinase-fragment complexes. These crystal fragments were classified by their binding cleft and subpockets, and their 3D structure and interactions were displayed in KinaFrag. In addition, the structural information, physicochemical information, similarity information, and substructure relationship information were contained in KinaFrag. Moreover, a computational fragment growing strategy obviously developed by our group was implemented in the KinaFrag. We test this fragment growing strategy using our fragment libraries, and obtained a lead compound of c-Met with ~1000-fold in vitro activity improvement compared with the hit compound. We hope KinaFrag could become a powerful tool for the fragment-based kinase inhibitor design. KinaFrag is freely available at http://chemyang.ccnu.edu.cn/ccb/database/KinaFrag/.