D-peptide binders are promising drug candidates that may offer better binding specificity and improved metabolic stability than canonical peptide drugs. However, there is a lack of efficient computational methods for the de novo design of D-peptide binders based on target protein structure. We developed a general framework for de novo design of D-helical peptide binders for the target protein, which consists of curved helical scaffold generation, scaffold docking to the target surface, Rosetta based sequence design, and in silico selection. For the convenience of conformation sampling, the targeted protein is mirrored to D-type, while the peptide ligands are presented in L-type during the sequence design step. We have applied this workflow to design D-helical peptides targeting the two major targets for inhibiting SARS-CoV-2, the receptor binding domain (RBD) of the spike protein and the main protease (3CLpro), to alter its oligomeric state and inhibit its activity. We found that both the receptor binding surface of RBD and the groove between the catalytic and regulation domains of 3CLpro are favorable for the binding of 28-mer D-helical peptides. We designed and tested 8 D-peptides for RBD and found 4 of them bound to RBD with the best one demonstrating submicromolar dissociation constant and the ability to block the binding of full-length spike protein toward its receptor, human angiotensin-converting enzyme 2. For 3CLpro, 3 of the 12 designed D-peptides could inhibit its catalytic activity. And the best peptide LY09 binds 3CLpro with submicromolar dissociation constant and disrupts the dimerization of 3CLpro. The D-peptide binder docking and design tools are publicly available at https://github.com/laiyii/D-peptide-binder-design.