Assessing the mechanical properties of recycled aggregate concrete (RAC) remains challenging due to the inherent heterogeneity of recycled aggregates and the complexity of the interfacial transition zones (ITZ). Although Scanning Electron Microscopy (SEM) has been widely used to investigate the microstructure of conventional concrete, its applicability for evaluating the mechanical performance of RAC has not been sufficiently examined. This study investigates whether SEM based microstructural descriptors can be used to estimate the mechanical properties of RAC. Six microstructural descriptors—including microstructural density, ITZ thickness, fiber engagement ratio, recycled pore filling, crack network tortuosity, and Particle Bound Index—were quantified from SEM images and integrated into a Microstructural Performance Number (MPN). An experimental database consisting of 48 mix designs incorporating different levels of recycled aggregates (0–100%), basalt fibers (0–2%), and binder compositions (plain cement, silica fume, fly ash, and combined SF+FA) was used to establish relationships between the microstructural indicators and the macroscopic mechanical strengths of concrete. Non linear exponential regression models were developed to relate the MPN to compressive, tensile, and flexural strengths. The proposed models showed coefficients of determination (R²) of 0.92, 0.88, and 0.89, respectively. The results indicate that SEM derived microstructural descriptors can provide a meaningful link between microstructural characteristics and the mechanical performance of RAC, offering a micro destructive approach for strength estimation. The findings further demonstrate that, within the proposed SEM based framework, an optimal RAC composition can be identified. Mixtures incorporating approximately 50% recycled aggregates, 0.5–1% basalt fibers, and combined silica fume–fly ash exhibited the highest microstructural performance, indicating the framework’s capability not only for strength assessment but also for microstructure-informed mix optimization.