Sustainable school improvement seems characteristic of successful schools (Stoll, Bolam, McMahon, Wallace & Thomas, 2006; Huffman, Olivier, Wang, Chen, Hairon & Pang, 2016). Successful schools can be defined as schools that educate students who are able to participate meaningfully in public life, there by forming a society characterised by healthy, inclusive democracy (Mahon, Hekkinen & Huttunen, 2019). These schools  have long experience of collaborative work with a focus on improving teaching and learning. During time they have changed the preconditions - the practice architectures – in such a way that they enables a professional learning community. On the other hand unsuccessful schools struggle with finding the right actions to get real improvement come around that affect the learning of the students. A theme in school improvement research is to study the actions of successful schools and accordingly, advice unsuccessful schools or school leaders to take the same actions (Leithwood, Harris & Hopkins, 2019). The objective of this study is to further examine the relationship between actions for sustainable improvement and practice architectures.We do that by returning to a study of a Swedish municipality that performed an extensive school improvement project implementing change agents the years 2009 to 2011. By letting teachers take the role of change agents the preconditions was altered in relation to the role of the principals. And by letting change agents lead professional learning team work new improvement actions were introduced. A change agent’s task is according to Schmuck and Runkel (1994) to design the process of organisational development. The teachers that was selected to be change agents were chosen based on their positive approach to school development. Furthermore several of them were defined as informal leaders in their schools. Within the framework of the project these teachers got an in-service training and at the same time they got an assignment to lead their colleges’ professional learning. Together with the principal, they were responsible for leading their schools development processes (Blossing, 2016).The study use theory of practice architectures (Kemmis and Grootenboer, 2008; Kemmis Wilkinson, Edward-Groves, Hardy and Grootenboer, 2014) in the framing of the research and the analysis of the research data. According to the theory a practice is understood as a socially established cooperative human activity, composed of the sayings, doings and relatings that hang together in a joint project. Different practices are characterised by different discourses (sayings) including the language used to describe, interpret, explain, orientate and justify a specific practice. Furthermore, practices are characterised by how they interact (relatings) among themselves and how they relate to other practices and various artefacts within and without the particular practice. In the same way, there are characteristic actions (doings) that are linked to a practice. Practice architectures constitute the enabling and constraining preconditions for the conduct of a specific practice and they appear in three intersubjective dimensions: the semantic, the physical and the social. In the semantic dimension, cultural-discursive arrangements appear through the language and speech surrounding the specific practice. In the social dimension, social-political arrangements reveal how people relate to each other as well as to artefacts inside and outside the practice. In the physical dimension, material-economic arrangements become visible in the actions and work that take place. Our research questions in this longitudinal study are as follows:1.       What arrangements and actions enables and constrains sustainable school improvement? 2.       In what way has the prevailing practice architectures affected the role of the change agent?