oftware systems become increasingly interconnected and complex, leading to a heterogeneous system landscape. This entails that architecture information and architecture documentation become more important. Currently, architecture documentation is a mostly manual task, which is costly, tedious and error prone. Even if initial documentation of a system’s architecture is available, it needs to be kept up-to-date as the system evolves, as otherwise its quality will decay to a point where it does not reflect the actual system and is not useful anymore. Therefore, automated support for maintaining and evolving architecture information and documentation of complex systems is highly beneficial to architects and other stakeholders. To achieve this, architecture information must be automatically recovered from heterogeneous data sources at different points in time and consolidated and integrated to provide an up-to-date representation of the system. Subsequently, the recovered architecture information must be automatically updated whenever data sources change over time. In this work, we present an early concept of a co-evolving digital architecture twin to model the system architecture via an architecture information model that combines and relates architecture information recovered from different sources at different points in time. We propose a framework for automated recovery, integration, and co-evolution of architecture information to create and maintain a digital architecture twin that is continuously and automatically updated as the system evolves. We present the general concepts and framework and discuss use cases to motivate benefits.