Research pairing ecological momentary assessment (EMA) methodology and ambulatory cortisol during daily life is still rare, as is careful testing of the within-person associations between stress, affect, and cortisol. Using a circumplex approach, we considered both valence and arousal components of affect. To examine the within-person covariation of momentary cortisol with momentary perceived stress, affective valence, and affective arousal in everyday life. 115 working adults (Mage = 41.2; 76% women; 76% white) completed six EMA surveys per day over 3 days. Each assessment included reports of perceived stress and affect (used to construct indicators of affective valence and arousal), followed by a saliva sample (from which cortisol was assessed). Multi-level models were used to examine the momentary associations between perceived stress, affective valence, affective arousal, and cortisol. Moments characterized by higher perceived stress were associated with higher cortisol (p = .036). Affective valence covaried with cortisol (p = .003) such that more positive valence was associated with lower cortisol and more negative valence with higher cortisol. Momentary affective arousal was not related to cortisol (p = .131). When all predictors were tested in the same model, only valence remained a significant predictor of cortisol (p = .047). Momentary perceived stress and affective valence, but not affective arousal, were associated with naturalistic cortisol. Cortisol was more robustly associated with affective valence than perceived stress or affective arousal. These findings extend our understanding of how moments of stress and particular characteristics of affective states (i.e., valence but not arousal) may "get under the skin" in daily life.