Trust in Web services will be greatly enhanced if these are subject to run-time verification, even if they were previously tested, since their context of execution is subject to continuous change; and services may also be upgraded without notifying their consumers in advance. Conversational Web services introduce added complexity when it comes to runtime verification, since they follow a conversation protocol and they have a state bound to the session of each consumer accessing them. Furthermore, conversational Web services have different policies on how they maintain their state. Access to states can be private or shared; and states may be transient or persistent. These differences must be taken into account when building a scalable architecture for run-time verification through monitoring. This paper, building on a previously proposed theoretical framework for run-time verification of conversational Web services, presents the design, implementation and validation of a novel run-time monitoring architecture for conversational services, which aims to provide a holistic monitoring framework enabling the integration of different verification tools. The architecture is validated by running a sequence of test scenarios, based on a realistic example. The experimental results revealed that the monitoring activities have a tolerable overhead on the operation of a Web service.