We define a flexible process as one that can change its behavior dynamically according to vari-able execution contexts; an adaptive process is one that can execute a service when conditions at runtime differ from those assumed during the ser-vice's initial design. While researchers have pro-posed several adaptation mechanisms, none of the existing frameworks systematically couples adap-tation design-time and runtime execution. Focus-ing only on design-time issues reduces an applica-tion's ability to adapt its behavior at runtime. In a loosely coupled environment, in fact, the actual execution context might differ dramatically from the conditions hypothesized during application design. In addition, implementing advanced process-execution mechanisms, such as service substitution, is possible only when processes and services are carefully designed in advance. To address this, we developed PAWS (Pro-cesses with Adaptive Web Services), a frame-work for flexible and adaptive execution of managed service-based processes. Our frame-work coherently supports both process design and execution. It also integrates several research results developed at Politecnico di Milano that address different aspects of adaptation, coupling design-time and runtime mechanisms in a global environment. We have two primary goals for PAWS. First, we want it be self-optimizing. PAWS should se-lect the best available services for executing the process and define the most appropriate qual-ity-of-service (QoS) levels for delivering them. Second, PAWS should guarantee service provi-sioning, even in case of failures, through re-covery actions and self-adaptation if the con-text changes. To meet these goals, PAWS provides methods and a toolset to support de-sign-time specification of all information re-quired for automatic runtime adaptation of focus