Knowledge of the fauna of soil mites in Brazilian soils is still scarce and limited to a few regions, referring almost exclusively to areas of natural vegetation. The purpose of this research was to evaluate the abundance and diversity of edaphic Gamasina mites (Mesostigmata) on ecosystems of natural vegetation and cultivations of prickly pear, cassava and pasture located in the Caatinga biome in the state of Alagoas, northeast Brazil. Collections were carried out every three months between December 2018 and August 2019. In each collection date and in each ecosystem, 16 soil samples were collected. During the study, a total of 2,979 Gamasina were collected. Representatives of 10 families were identified, one of which, new to science. The largest number of Gamasina collected belonged to Ascidae (60%), followed by Laelapidae (14.3%) and Rhodacaridae (10%). Forty-four morphospecies were found, of which 25 new to science, distributed in 22 genera, two new to science. Protogamasellus mica Athias-Henriot, Protogamasellus sigillophorus Mineiro, Lindquist & Moraes and a new species of Protogamasellus (Ascidae) were the most abundant species. Mean abundance and richness were highest in the natural vegetation (respectively 38.2 mites and 12.6 species per sample), followed by prickly pear (29.6 and 8.6), pasture (20.8 and 8.6) and cassava (10.9 and 5.7). The Shannon-Weaver's index diversity was highest for pasture (2.6) and lowest for the natural vegetation (2.2). The equitability indexes were quite similar in the different ecosystems though slightly lower in the natural vegetation (0.7 in the natural vegetation and 0.8 in other ecosystems). A cluster analysis indicated that the area of natural vegetation was the most distant from the others. Several of the species collected belong to Laelapidae and Macrochelidae, families with species used in the applied biological pest control. Thus, it is hoped that this work will serve as a basis for future applied research, aimed at prospecting Gamasina potentially useful for use in biological control programs.