This work's starting point is the observation of the heterogeneity of algorithms and data sources involved in Augmented Reality (AR) applications. The idea is to be able to design AR applications taking advantage of modules offered as services from different developers and using data from different sources. In addition, an increasing number of AR applications is deployed within mobile devices and involve pervasive computing. Thus, a framework aiming to allow developers to create state of the art applications should also offer the built-in ability to develop transparently distributed applications and pervasive services. This paper presents the design of a framework to create AR applications using services and data from different sources in a transparent and efficient way. The goal of the framework is to offer ability to build composite applications combining locally running functionality and remote AR services. This framework is the result of the extension of a previously developed component based framework named ARCS (Augmented Reality Component System). With this extension, any component of ARCS (eg. feature detection, matching, rendering) becomes accessible to any non-ARCS application and any ARCS based application has transparent access to available web services and data sources. The framework design goals and assessment criteria are based on AR applications requirements defined through a dedicated methodology. Proof applications are also presented to show how the framework answers state of the art AR applications needs.