Agility, characterized by rapid whole-body movement in response to external stimuli, is a key performance determinant in team sports owing to its reliance on perception-action coupling. Despite its importance, existing evidence on training effects is disproportionately focused on change-of-direction ability, overlooking the perceptual-cognitive demands of true agility. This systematic review therefore sought to quantify the effects of agility-specific training, examine moderators and establish dose-response model in team-sport players, while accounting for perception-action coupling. Following PRISMA 2020 guidelines, systematic searches were conducted in Web of Science, Scopus and PubMed for peer-reviewed English-language studies. Risk of bias was appraised using a modified Cochrane Collaboration’s tool and study quality was evaluated via a tailored PEDro scale. Effect size (ES) was calculated using Hedge’s g, and dependencies among multiple ESs within studies were addressed through three-level meta-analysis. Subgroup analyses and both linear and non-linear meta-regressions were performed to examine potential moderators and establish dose-response models. Twenty-eight studies met the inclusion criteria, of which 26 contributed 53 ESs to quantitative analyses. Agility training produced a large improvement in reactive agility test (RAT) performance (ES = 0.65, p < 0.01), a moderate improvement in pre-planned agility test (PAT) performance (ES = 0.55, p < 0.01), and a non-significant moderate effect in reaction test (RT) outcomes (ES = 0.52, p > 0.05). Training effects were moderated by participants’ characteristics, with junior athletes (ES = 0.77, p < 0.05) and national level athletes (ES = 0.80, p < 0.05) demonstrating greater responses. However, female and mix-gender samples were underrepresented, and the evidence base was dominated by studies in soccer and basketball. Dose-response modelling revealed a curvilinear relationship between training duration and RAT (QM = 11.64, p < 0.01), peaking at 7-8 weeks and a positive linear association between training frequency and RAT (β = 0.172, p < 0.01). No significant relationship was observed between session time and RAT (p > 0.05), although most positive ESs clustered around 20-25 minutes per training session. Agility training exerts a large overall effect on RAT performance in team-sport players, with outcomes moderated by age and training status. Interventions of 7-8 weeks delivered at higher frequency (> 3 times/week) with 20-25 minutes session duration are frequently associated with favorable adaptions. These recommendations, however, should be interpreted cautiously given the moderate-to-low certainty of evidence, high within-study variability, dominance of soccer and basketball samples and potential risk of publication bias.