Heterogeneity in host and gut microbiota hampers microbial precision intervention of type 2 diabetes mellitus (T2DM). Here, we investigated novel features for patient stratification and bacterial modulators for intervention, using cross-sectional patient cohorts and animal experiments. We collected stool, blood, and urine samples from 103 patients with recent-onset T2DM and 25 healthy control subjects (HCs), performed gut microbial composition and metabolite profiling, and combined it with host transcriptome, metabolome, cytokine, and clinical data. Stool type (dry or loose stool), a feature of the stool microenvironment recently explored in microbiome studies, was used for stratification of patients with T2DM as it explained most of the variation in the multiomics data set among all clinical parameters in our covariate analysis. T2DM with dry stool (DM-DS) and loose stool (DM-LS) were clearly differentiated from HC and each other by LightGBM models, optimal among multiple machine learning models. Compared with DM-DS, DM-LS exhibited discordant gut microbial taxonomic and functional profiles, severe host metabolic disorder, and excessive insulin secretion. Further cross-measurement association analysis linked the differential microbial profiles, in particular Blautia abundances, to T2DM phenotypes in our stratified multiomics data set. Notably, oral supplementation of Blautia to T2DM mice induced inhibitory effects on lipid accumulation, weight gain, and blood glucose elevation with simultaneous modulation of gut bacterial composition, revealing the therapeutic potential of Blautia. Our study highlights the clinical implications of stool microenvironment stratification and Blautia supplementation in T2DM, offering promising prospects for microbial precision treatment of metabolic diseases. Article Highlights