The Internet of Things (IoT) connects everyday objects with each other and with the Internet. IoT applications are complex to develop. Model-driven methods promise to make the complex development of IoT applications manageable by raising the level of abstraction. Related work has proposed a variety of component and connector (C&C) architecture description languages (ADLs) for developing IoT applications. However, these mainly focus on the early stages of development and largely neglect reliability aspects. We present MontiThings, an ecosystem for model-driven IoT applications. MontiThings specifies an IoT-focused C&C ADL using the MontiCore language workbench. MontiThings covers the lifecycle of IoT applications, from initial architectural concepts to runtime deployment and analysis. MontiThings offers reliability mechanisms that can help specify resilient applications. In addition, MontiThings provides an integration method for hardware drivers that enables clear separation of concerns, enabling reuse and testing of components independent of their hardware integration. A generator translates the C&C architecture models into C++ code. Tagging can be used to integrate IoT components with synthesized digital twins. MontiThings' requirements-based deployment method is able not only to compute a distribution of components to IoT devices, but also to actively suggest changes to users if their requirements cannot be met. If devices fail at runtime, MontiThings automatically adapts the deployment to the changed situation. To understand unforeseen situations that may occur at runtime, MontiThings provides developers with model-driven analysis services. Overall, MontiThings demonstrates an end-to-end model-driven approach for designing IoT applications.