The CHARMM General Force Field (CGenFF) is a force field for molecular mechanics simulations of drug-like molecules. This article describes the algorithms for bond perception and atom typing in CGenFF. The atom type assigns attributes to atoms and bonds, such as valence, bond order, and ring membership. A programmable decision tree is used to assign CGenFF atom types based on these attributes. The atom typer was validated on 477 model compounds, including 126 test-set molecules. The algorithm involves ring perception, which identifies rings with less than 8 atoms. A two-stage ring perception algorithm was implemented, based on published work. The first stage reduces the molecule's connectivity to a Homeomorphically Reduced Graph (HRG), while the second stage uses a message-passing algorithm to identify rings. The algorithm also resolves resonance structures, assigning bond orders based on connectivity and aromaticity. The atom typing rules are based on a decision tree, allowing for flexible and efficient assignment of atom types. The algorithm was tested on various molecules, including the pyridinium ion and 1,3-dipentene, demonstrating its effectiveness in assigning atom types and bond orders. The results show that the CGenFF atom typer can accurately assign atom types and bond orders, making it a valuable tool for molecular mechanics simulations of drug-like molecules.The CHARMM General Force Field (CGenFF) is a force field for molecular mechanics simulations of drug-like molecules. This article describes the algorithms for bond perception and atom typing in CGenFF. The atom type assigns attributes to atoms and bonds, such as valence, bond order, and ring membership. A programmable decision tree is used to assign CGenFF atom types based on these attributes. The atom typer was validated on 477 model compounds, including 126 test-set molecules. The algorithm involves ring perception, which identifies rings with less than 8 atoms. A two-stage ring perception algorithm was implemented, based on published work. The first stage reduces the molecule's connectivity to a Homeomorphically Reduced Graph (HRG), while the second stage uses a message-passing algorithm to identify rings. The algorithm also resolves resonance structures, assigning bond orders based on connectivity and aromaticity. The atom typing rules are based on a decision tree, allowing for flexible and efficient assignment of atom types. The algorithm was tested on various molecules, including the pyridinium ion and 1,3-dipentene, demonstrating its effectiveness in assigning atom types and bond orders. The results show that the CGenFF atom typer can accurately assign atom types and bond orders, making it a valuable tool for molecular mechanics simulations of drug-like molecules.