An approach for providing voice processing over a packet switched network is disclosed. A communication system includes a special forking proxy server that receives a message for establishing a session between a calling user agent and a called user agent. The message has a first payload corresponding to a first call leg. The system also includes a bridging user agent that communicates with the first network element and generates a response that contains the first payload and a second payload that corresponds to a second call leg. The bridging user agent is configured to bridge the first call leg and the second call leg to establish a session between the calling user agent and the called user agent. The special forking proxy server is configured to fork an acknowledgement message from the calling party to acknowledge the session establishment by the called user agent. The present invention has particular applicability to SIP (Session Initiation Protocol) IP (Internet Protocol) telephony services.