Affibody is a class of smallest antibodies consisted of 58 amino acids, of which 13 amino acids are variable. Phage display is currently used for the production of affibody, but this method needs huge library construction for each antigen. In this study, we tried to develop the affibody against receptor binding domain (RBD) of SARS-CoV-2 spike protein in silico. SARS-CoV-2 virus binds to human ACE2 via spike protein, which is, therefore, a promising target for preventing SARS-CoV-2 infection. We analyzed three-dimensional coordinates of RBD structure and set antibody binding site on the interface surface between RBD and ACE2. We then designed many affibody structures based on the electrostatic potential of the binding site on RBD, and performed in silico docking prediction of constructed affibodies and RBD. Thereafter, two affibody molecules that showed the highest and second highest docking score as well as two molecules with much lower scores were expressed by wheat germ cell-free protein synthesis method. Binding affinity between expressed affibody molecules and RBD was examined with AlphaScreen assay, and two high score affibody molecules showed significantly higher luminescent signals due to binding than control low score molecules. In contrast, all four molecules showed no binding signals to negative control DHFR. The present study shows the successful non-animal-derived, cell-free and in silico generation of artificial antibody, targeting precise protein surface of SARS-CoV-2 RBD.