The UK Mental Capacity Act provides an important legislative framework for protecting persons who are vulnerable, by virtue of partial or total lack of capacity, from abuse, coercion, exploitation, disrespect or unwarranted intrusions on their privacy and liberty. In this article we argue, however, that in order to effectively implement the Act's governing principles and promote the autonomy of persons whose capacities are impaired, health and social welfare professionals must go beyond a primarily cognitive approach to capacity assessment and refer to a range of more demanding autonomy conditions relating to authenticity, diachronic coherence and consistency, accountability to others, and affective attitudes towards oneself. We suggest that these conditions are in fact implicit in the guidelines, scenarios and practical advice contained in the Act's accompanying Code of Practice. To ground our analysis, we discuss two cases of end-of-life decision-making, and consider how these cases might have been assessed had the Mental Capacity Act been applied. We also address the question of whether the conditions for autonomy we identify raise the bar too high to be met by persons who are vulnerable because their capacity is partially compromised.