Off-grid communities in Kenya face challenges in accessing reliable energy sources for agricultural productivity. A mixed-methods approach was employed, including surveys, interviews, and data from monitoring devices. Multilevel regression models were used to analyse the data at both individual household and community levels. The average system reliability score for off-grid communities in this study was 82%, with a confidence interval of ±5%. Multilevel regression analysis revealed that access to clean water significantly improved system reliability by 10 percentage points, while high temperatures reduced it by 6 percentage points. Investment should be directed towards improving water supply systems and building climate-resilient infrastructure in off-grid communities. Off-Grid Communities, Multilevel Regression Analysis, System Reliability, Agricultural Productivity The empirical specification follows $Y=\beta_0+\beta^\top X+\varepsilon$, and inference is reported with uncertainty-aware statistical criteria.