Introduction The high prevalence of falls, lack of stability and balance, and general physical deconditioning are concerning issues for longevity and quality of life for adults aged 65 years and older. Although supervised delivery of the Otago Exercise Program (OEP) has demonstrated evidence of effectiveness in reducing fall risk of older adults, opportunities for ongoing unsupervised exercise performance are warranted. An option to facilitate exercise and performance of health behaviors may be via a social robot. The purpose of this study was to examine feasibility and initial outcomes of a robot-delivered fall prevention exercise program for community-dwelling older adults. Methods Five participants aged 65 years and older were recruited to receive robot-delivered modified OEP and walking program three times per week for four weeks. Outcomes of demographics, self-reported performance measures (Modified Falls Self-Efficacy Scale, Activities-specific Balance Confidence, and Almere Model assessing various constructs of acceptance of use of robotic technology), and physical performance measures (Timed Up and Go Test, Short Physical Performance Battery, Balance Tracking System [BTrackS] center of pressure sway) were collected. Data were analyzed descriptively and examined for trends in change. Measures of central tendency and distribution were used according to the distribution of the data. Results The mean age of the participants was 75 years (range: 66-83 years; four females and one male). The range of participant exercise session completion was 7-12 (mode=11, n=3). Constructs on the Almere Model that started and remained positive were Attitudes Toward Technology and Perceived Enjoyment with the robot. Anxiety improved from 3.80 to 4.68, while Social Presence of the robot improved from 2.80 to 3.56. The construct of Trust was somewhat negative among participants upon commencing the program and did not substantially change over time. Two participants improved their confidence on the Activities-specific Balance Confidence scale by more than 10%, while all participants showed some improvement in confidence in their balance. Mixed results were found with the Modified Falls Self-Efficacy Scale. Mean gait speed for the participants improved by 0.76 seconds over 3 meters. Improvement was also demonstrated for the Short Physical Performance Battery, with two participants improving scores by 2-3 points out of 12. No appreciable changes were found with the Timed Up and Go test and the BTrackS assessment. Conclusion Using a robot-led exercise program is an accessible and feasible way to deliver exercise to community-dwelling older adults in the home, but some technical constraints remain. Outcomes suggest that a four-week program is sufficient to elicit some positive trends in health outcomes and has the potential to reduce fall risk.