Eukaryotic transcriptomes and proteomes are immensely complex: many genes express alternative RNA isoforms, which can be translated into multiple, functionally distinct, proteoforms. Uncovering the functions of different RNA and protein isoforms requires an understanding of which cell types and stages they are expressed in, but there are currently no tools to integrate transcriptomic and proteomic results together in a single-cell context. Furthermore, many proteins and proteoforms remain uncharacterised because proteomic techniques can only detect the proteins present in incomplete reference databases. To address these issues, we developed GenomeProtSC, a user-friendly R Shiny webserver for long-read single-cell proteogenomics. GenomeProtSC can identify expressed genes and isoforms from long-read single-cell RNA-seq data and use this to generate a sample-specific proteome database. It can then search mass spectrometry peptide data against the generated database, revealing the translation of known and novel protein-coding RNA isoforms with cell-type resolution. We demonstrated GenomeProtSC on long-read single-cell RNA-seq data and mass spectrometry data from 1-, 3- and 6-month human cortical organoids: it identified 127,625 unique peptides; 10,133 translated genes; and 4,311 known and 501 novel translated open reading frames (ORFs). Novel ORFs supported by uniquely mapped peptides were shorter than known ORFs and included the translation of long non-coding RNAs, novel proteoforms, pseudogenes and retained introns. We discovered the cell-type-specific expression and translation of genes with unknown function, such as LRRC3B, which is enriched in neural progenitors and expressed six RNA isoforms, four of which are protein-coding and encode a translated ORF. GenomeProtSC enables researchers to elucidate the transcriptomes and proteomes of individual cell types and states, putting the power of proteogenomics into the hands of both bioinformaticians and biologists.