Abstract Green roofs are promoted to contribute to more resilient cities by enhancing urban ecosystem services and food systems. Extensive, low‐maintenance green roofs experience frequent environmental stresses, which reduce plant survival and growth. Stress‐tolerant plants are therefore used to sustain well‐established services, such as building temperature regulation. However, transitioning extensive green roofs to provide other key urban services, such as food production, involves less tolerant plant species. Although facilitation exerted by stress‐tolerant species (nurses) has been proposed to improve the performance of stress‐intolerant species (protégés) in extensive green roofs, the conditions under which facilitation could occur are not well understood. Therefore, a comprehensive framework is needed that integrates current knowledge on how the performance of protégé species is affected by nurse plants across stress conditions. We present a framework for green roof research that results in a linear model that integrates (i) modern trait–environment theory and (ii) facilitation ecology in a refined stress‐gradient hypothesis (SGH) originally developed following study of other stressful environments. The model makes testable predictions on how phenotypic traits mediate the performance response of protégé species to nurse plants along stress gradients in extensive green roofs. This is not only useful for the analysis of eco‐physiological performance measures directly linked with multifunctionality and ecosystem services, but also demographic or ‘vital’ rates that drive species persistence and plant community maintenance. We discuss a range of applications related to key agricultural and ecological questions arising from contemporary extensive green roof research, such as enhancing conditions for crop production, weed management, plant invasions and biodiversity conservation. We also provide guidelines for the generation of appropriate data and for fitting this model using readily available statistical procedures. Our framework will allow researchers to assess under which environmental conditions nurse–protégé interactions are feasible. We expect the findings from such research to help develop strategies and guidelines for managing environmental conditions that optimize protégé performances that ultimately affect the delivery of ecosystem services in constructed urban green spaces.