The evaluation of cognitive status is not routine in the acute stroke setting. We aimed to investigate feasibility, applicability, and performances of the Montreal cognitive assessment (MoCA) in acute stroke patients. Consecutive stroke patients (ischemic or hemorrhagic) admitted to one stroke unit were evaluated 5-9 days after stroke with MoCA (score range: 0-30; higher scores indicate better cognitive performance). Pre-morbid functional and cognitive status was assessed by a structured interview to caregivers. Neuroimaging data regarding index stroke and pre-existing lesions were collected. From December 2009 to December 2010, out of 207 patients with stroke, 137 (66%) were enrolled [mean age 69.2 ± 14.8 years; males 62%; mean National Institute of Health and Stroke Scale (NIHSS) score 5.9 ± 7.9]. The most common reason for non-enrolment was unfitting the time window inclusion criteria. MoCA was entirely applicable to 113/137 (82.5%) patients and the mean score was 17.8 ± 7.1. Multivariate analyses showed that non-applicability was associated with higher NIHSS scores [OR (95% CI) = 1.4 (1.2-1.7) for each point], left sided lesions [OR (95% CI) = 18.8 (2.3-155.2)], and worse pre-morbid functional status [OR (95% CI) = 0.7 (0.6-0.9) for each point of the instrumental activity of daily living scale]. Factors influencing MoCA performance were low education (β = 0.264, p < 0.01), higher NIHSS scores (β = -0.277, p < 0.01) and worse pre-morbid functional status (β = 0.504, p < 0.001). MoCA administration is feasible in acute patients with mild-to-moderate stroke, with lesion location, stroke severity, and pre-morbid functional status as major determinants of its applicability and performance. MoCA seems to reveal some degree of cognitive deficit even in patients with mild stroke.