Contemporary neurocognitive models of drug addiction have associated this condition with changes in interoception —namely, the sensing and processing of body signals that fulfill homeostatic functions relevant for the onset and maintenance of addictive behavior. However, most previous evidence is inconsistent, behaviorally unspecific, and virtually null in terms of direct electrophysiological and multimodal markers. To circumvent these limitations, we conducted the first assessment of the relation between cardiac interoception and smoked cocaine dependence (SCD) in a sample of (a) 25 participants who fulfilled criteria for dependence on such a drug, (b) 22 participants addicted to insufflated clorhidrate cocaine (only for behavioral assessment), and (c) 25 healthy controls matched by age, gender, education, and socioeconomic status. We use a validated heartbeat-detection (HBD) task and measured modulations of the heart-evoked potential (HEP) during interoceptive accuracy and interoceptive learning conditions. We complemented this behavioral and electrophysiological data with offline structural (MRI) and functional connectivity (fMRI) analysis of the main interoceptive hubs. HBD and HEP results convergently showed that SCD subjects presented ongoing psychophysiological measures of enhanced interoceptive accuracy. This pattern was associated with a structural and functional tuning of interoceptive networks (reduced volume and specialized network segregation). Taken together, our findings provide the first evidence of an association between cardiac interoception and smoked cocaine, partially supporting models that propose hyper-interoception as a key aspect of addiction. More generally, our study shows that multimodal assessments of interoception could substantially inform the clinical and neurocognitive characterization of psychophysiological and neurocognitive adaptations triggered by addiction.