The presence of animal grazing at different intensities in integrated crop livestock systems (ICLS) may affect the forms and the balance of K in the soil, especially after many years. The objective of this study was to evaluate different K forms and their balance for 17 years with ICLS. ICLS experimental area used no-till practices with different grazing intensities of cattle in winter pasture (black oat + annual ryegrass) and soybean production in the summer season. The experiment started in May 2001 at a location with a Rhodic Hapludox (Oxisol). The treatments consisted of varied grazing intensities: intensive grazing (IG) at 10 cm of pasture height, moderate grazing (MG) at 20 cm of pasture height, and a no grazing (NG) treatment. The experimental design was a randomized block design with three replicates. Soil samples were collected in 2001, 2007, 2012 and 2017 in 0−5, 5−10, 10−15 and 15−20 cm of soil layers. Desorption of K was measured using the Mehlich-1 method and the results were fitted to first-order kinetic equations to obtain readily and potentially available K content. Total K was determined by X-ray fluorescence, and the unavailable K by the difference between the total K and potentially available K. Long-term no-till ICLS resulted in an increase in more readily available K forms in the soil surface. The increase was especially prominent in the no-grazing treatment due to the greater amount of K left in the plant residues. Regardless of grazing intensity, soybean cultivation was the controlling component of the soil nutrient budget, representing 98.6 % of the total amount of K exported. Animal outputs represent only 1.4 % of total K exported and are therefore the recycling component of the system. Although meat production in ICLS represents only a very small fraction of the amount of K exported from the system, K tends to become less available in the grazed treatments in the long-term.