In this contribution, we present a framework for an intelligent natural language interface (NLI) that suits the need of embedded platform, using agent-based approach. The proposed framework is motivated by an ongoing speech technology research project aimed at developing a generic synthesizer for information disseminating systems in local languages. The architecture is based on various forms of action representations with a sequence of transformations that converts users’ input (text or speech) into a suitable set of agent actions that produce response to the input. This approach incrementally minimizes the complexity and ambiguity of the natural language input by using predefined sets of interim actions at different levels, hence, increasing the robustness and reliability of the NLI.