Developing effective water treatment technologies is indispensable to mitigate the negative impacts that aqueous contaminants can have on the environment. In this work, the potential of agro-industrial residues such as peanut shells, rice husk, coconut fiber and corn cob was investigated for removing caffeine from water using fixed-bed columns. The effects of the following parameters on the adsorptive performance of the columns were studied: particle size of the residues (Ps = 595–710, 710–1400, 1400–2000 μm), bed height (Z = 10, 15, 20 cm), flow rate (Q = 0.5, 1.0, 1.5 mL/min), and caffeine concentration (Co = 5, 10, 15 mg/L). Results demonstrated that the effect of particle size was influenced by the bulk density of the material (ρ) and bed porosity (ԑ). Increasing the bed height and decreasing the flow rate led to longer residence times and higher adsorption capacities. Increasing the caffeine influent concentration, increased the adsorption capacity, but reduced the removal efficiency. Columns packed with coconut fiber, Ps = 710–1400 μm (ρ = 0.102, ԑ = 0.30), Z = 20 cm, Q = 0.5 mL/min, and Co = 15 mg/L resulted in the best adsorption capacity (0.326 mg/g) and removal efficiency (47 %). The Adams-Bohart, Thomas, Yoon-Nelson, and Dose-Response models were used to describe the dynamic adsorption of the columns. Thomas' model predicted adsorption capacities closer to the experimental ones (R2 = 0.996, for 0.326 mg/g). The results obtained also open the possibility of defining criteria for assessing materials for removing aqueous pollutants in continuous processes.