STEP-NC (ISO 14649) is a newly developed and feature-based machine tool control language for CNC controller. In contrast with conventional G&M code (ISO 6983), this new data model could almost transmit all the designing and planning information for machining parts through CAD&CAPP&CAM to CNC. However, STEP-NC is feasible but not enough to support some intelligent actions such as self-planning, self-learning, decision-making, etc. A complete and well-structured knowledge base is urgently needed to be established to support CNC intelligence. The main work of this paper is focused on two fields: i) An extraction, presentation and utilization mechanism for ontology-based machining knowledge is implemented. ii) A novel framework for STEP-Compliant and knowledge-base supported intelligent CNC controller is proposed.