The reduced solubility of some active pharmaceutical ingredients (APIs) in water is a limiting factor for the development of new formulations. Therefore, the search of new compatible solvents would be convenient to overcome such limitation. Among this group, eutectic solvents stand out as good candidates. In this work, the systematic application of COSMO-RS as a predictive method to anticipate APIs solubility in systems involving natural eutectic solvents (NAESs) is presented. The methodology has been validated by using a collection of 50 APIs – NAESs systems from the literature. This predictive tool allowed the selection of four NAESs (reported in literature as natural eutectic mixtures) as good candidates to improve lidocaine and caffeine solubility in water and PEG 300 (benchmark solvents). The four NAESs candidates were prepared and experimentally characterized by means of density and viscosity measurements. Density values ranging from 1.0 to 1.3 g/mL and viscosities from 200 to 2,200 cP were found, with glycerol based NAESs as the most viscous systems. Solubility experiments were also carried out to analyze COSMO-RS predictions validity. Betaine: lactic acid and betaine: 1,2-propanediol are good candidates for lidocaine solutions, achieving 2,300 and 572 mg/mL values respectively, much higher than polyethylene glycol, PEG 300, (173 mg/mL), typically used as benchmark solvent. In the case of caffeine, proline:lactic acid was found as the best candidate with 30 mg/mL solubility (higher than water: 23 mg/mL and PEG 300: 18 mg/L). This work permits a systematic analysis of APIs solubility in NAESs (more than 80 systems reported) by using a total predictive tool (COSMO-RS predictions) to anticipate that promising solvents, with high solubilities for lidocaine and caffeine.