Multi-view clustering leverages the semantic information of data from different views to achieve superior performance. Recent works often focus on fusing multi-view information through unidirectional alignment from local to global levels, emphasizing coarse-grained global feature alignment and inter-view distribution matching. However, this approach overlooks fine-grained clustering semantics and view-specific differences, making it challenging to achieve consistency in multi-view alignment. To address these limitations, we propose a novel method named Hierarchical Prompt-guided alignment for Multi-View Clustering (HiPMVC). To enable the extraction of fine-grained clustering semantics, we introduce prompt learning into the multi-view clustering task with hierarchical design, addressing heterogeneous information at both the primary and embedding levels. Furthermore, to address the challenge of achieving consistency across views, we propose a bidirectional prompt-guided alignment strategy, ensuring more stable and globally consistent distribution alignment. Our method is validated through experiments conducted on seven datasets, showcasing its effectiveness.