The phylogenetic history of termites has been investigated using mitochondrial genomes and transcriptomes. However, both sets of markers have specific limitations. Mitochondrial genomes represent a single genetic marker likely to yield phylogenetic trees presenting incongruences with species trees, and transcriptomes can only be obtained from well-preserved samples. In contrast, ultraconserved elements (UCEs) include a great many independent markers that can be retrieved from poorly preserved samples. Here, we designed termite-specific baits targeting 50,616 UCE loci. We tested our UCE bait set on 42 samples of termites and three samples of Cryptocercus, for which we generated low-coverage highly-fragmented genome assemblies and successfully extracted in silico between 3,426 to 42,860 non-duplicated UCEs per sample. Our maximum likelihood phylogenetic tree, reconstructed using the 5,934 UCE loci retrieved from upward of 75% of samples, was congruent with transcriptome-based phylogenies, demonstrating that our UCE bait set is reliable and phylogenetically informative. Combined with non-destructive DNA extraction protocols, our UCE bait set provides the tool needed to carry out a global taxonomic revision of termites based on poorly preserved specimens such as old museum samples. The Termite UCE database is maintained at: https://github.com/oist/TER-UCE-DB/.