Chemical modification of transcripts with 5′ caps occurs in all organisms. Here, we report a systems-level mass spectrometry-based technique, CapQuant, for quantitative analysis of an organism's cap epitranscriptome. The method was piloted with 21 canonical caps—m 7GpppN, m 7GpppNm, GpppN, GpppNm, and m 2,2,7GpppG—and 5 ‘metabolite’ caps—NAD, FAD, UDP-Glc, UDP-GlcNAc, and dpCoA. Applying CapQuant to RNA from purified dengue virus, Escherichia coli, yeast, mouse tissues, and human cells, we discovered new cap structures in humans and mice (FAD, UDP-Glc, UDP-GlcNAc, and m 7Gpppm 6A), cell- and tissue-specific variations in cap methylation, and high proportions of caps lacking 2′- O-methylation (m 7Gpppm 6A in mammals, m 7GpppA in dengue virus). While substantial Dimroth-induced loss of m 1A and m 1Am arose with specific RNA processing conditions, human lymphoblast cells showed no detectable m 1A or m 1Am in caps. CapQuant accurately captured the preference for purine nucleotides at eukaryotic transcription start sites and the correlation between metabolite levels and metabolite caps.