In this paper, we introduce a semantically enriched capability matching model for agent services. Our vision is to integrate both agent and semantic web services and provide the interoperability of agents under the semantic web extension. In multi agent system architectures, there is a specific agent or service called directory facilitator which is responsible to keep knowledge about the services given by the agents within the system. Other agents query in directory facilitator to identify agents that provide the required services. Hence, automated service discovery in multi agent systems is a critical issue. Here, we propose a matching engine architecture in which capabilities of agent services are handled semantically and it replies agent service requests with most suitable service advertisements. The paper includes formal basics and design details of this engine and also discusses its implementation with a proper case study.