Investigating the multimolecule patterns in living cells is of vital importance for clinical and biomedical studies. Herein, we reported for the first time the engineering of gold nanoflares as smart automata to implement computing-based diagnosis in living mammalian cells. Defining the logic combinations of miR122 and miR21 as the detection patterns, the corresponding OR and AND diagnostic automata were designed. The results showed that they could recognize the correct patterns rapidly and sensitively. The automata could enter cells via self-delivery and have good biocompatibility. They enabled accurate diagnosis on miRNA signatures in different cell lines and differentiation of fluctuations in the same cell line at single cell resolution. Moreover, the automata afforded an innovative diagnostic mode. It simplified the complicated process of detecting, data-collecting, computing, and evaluating. The direct diagnosing result ("1" or "0") was exported according to the embedded computation code. It highlighted the new possibility of using smart automata for intelligent diagnostics and cancer therapy at single cell resolution.