Many real-time distributed collaborative applications are emerging that require exchange of critical sensor data among geographically distant end users under resource-constrained network conditions. The QoS requirements, e.g., required bandwidth, latency, acceptable data quality, and reliability are interdependent, and critical to the operation of these applications. This paper presents an AWON (application-aware overlay networks) architecture for deploying application-aware services in an overlay network to best meet the application requirements over the available overlay networking infrastructure. An application programming interface (API) is presented to facilitate development of applications within the AWON architectural framework. The API supports the configuration of overlay nodes for in-network, application-aware processing. Application-defined plug-in modules are used to deploy application-specific functionality at each overlay node. The API also enables communication between application and the overlay routing protocol for the desired QoS support. The effectiveness of the AWON architecture and the API is demonstrated for a real-time weather radar data dissemination application using planetlab. Experimental results show that AWON-based application-aware services significantly improve the quality of the content delivered to the end users in bandwidth-constrained conditions.