Recent analyses of bacterial genomes have shown that integrated elements that transfer by conjugation play an essential role in horizontal gene transfer. Among these elements, the integrative and mobilizable elements (IMEs) are known to encode their own excision and integration machinery, and to carry all the sequences or genes necessary to hijack the mating pore of a conjugative element for their own transfer. However, knowledge of their prevalence and diversity is still severely lacking. In this work, an extensive analysis of 124 genomes from 27 species of Streptococcus reveals 144 IMEs. These IMEs encode either tyrosine or serine integrases. The identification of IME boundaries shows that 141 are specifically integrated in 17 target sites. The IME-encoded relaxases belong to nine superfamilies, among which four are previously unknown in any mobilizable or conjugative element. A total of 118 IMEs are found to encode a non-canonical relaxase related to rolling circle replication initiators (belonging to the four novel families or to MobT). Surprisingly, among these, 83 encode a TcpA protein (i.e., a non-canonical coupling protein (CP) that is more closely related to FtsK than VirD4) that was not previously known to be encoded by mobilizable elements. Phylogenetic analyses reveal not only many integration/excision module replacements but also losses, acquisitions or replacements of TcpA genes between IMEs. This glimpse into the still poorly known world of IMEs reveals that mobilizable elements have a very high prevalence. Their diversity is even greater than expected, with most encoding a CP and/or a non-canonical relaxase.