Several trials were conducted to study the persistence and productivity of multiple mixtures of creeping legumes associated with grasses during and after the establishment in animal grazing. A random block design with five replicates was used for each of the associations under study. The treatments were two grazing intensities of 100 and 200 LAU/ha. The legumes were: Centrosema pubescens (centro), Macroptilium atropurpureum (siratro), Pueraria phaseloides (tropical kudzu), Neonotonia wightii (glycine), Stilosantes guyanesis (style) and Teramnus labialis (teramnus) associated with Panicum maximum (Guinea grass), Brothriochloa pertusa (Indian bluegrass), Paspalum notatum (bahia grass), Brachiaria decumbens (signal grass) and Dichanthium caricosum (nadi bluegrass) and they sown on a red or brown ferralic soil on carbonate at the beginning of the rainy season. The combination of mechanical labors such as the cultivation and the pass of the cutter diminished the weeds. When using animals, weeds decreased from 15 to 1 %, which proves that the area was well managed and there is adequate control. The lower grazing intensity, 100 LAU/ha, showed that its effect on the association (Guinea grass-legumes) improved its components in time, with absence of weeds and high percentage of legumes in the association (49.8 vs 50.7 % at the beginning and end of the trial, respectively). The same situation was found for the association with signal grass or natural grasslands. It is concluded that animals may be used for the successful establishment of legumes when they are used between 30-40 per hectare, short occupation time and flexible recovery time, as a method for weeds control. Grazing intensities of 100 LAU/ha allowed good development and productivity of the association grasses-multiple mixture of legumes in time, whereas 200 diminished the components of the association and the weeds.