Antimicrobial peptides (AMPs) are ancient oligopeptides of various length, widespread in all living organisms. They participate in host defense against bacteria, enveloped viruses, fungi, and even cancer cells. AMPs do not display consensus sequences but do share some common features, such as positive charge, hydrophobicity, and amphipathicity. Their main antibacterial activity is associated with the disruption of the cell membrane.Due to the current drastic loss of antibiotic effectiveness, AMPs are an important alternative. Therefore, we developed AmpGram - a new tool for the identification of the putative AMPs without performing expensive experiments. Moreover, AmpGram is the first AMP prediction software suitable for the identification of cryptic AMPs, fragments possessing AMP properties within longer sequences. Therefore, it can be used as a proteome screening tool.AmpGram uses n-grams (subsequences) to reveal amino acid motifs associated with the presence or absence of antimicrobial properties. Prediction is performed using a stacked random forest model. AmpGram is available as a web server for multiple query sequences and as an R package intended for proteome screening.