This paper proposes a non-parametric graph clustering (NpGC) framework for multi-view data, aiming to eliminate hyper-parameters that complicate the clustering process. The framework uses two types of anchors—view-related and view-unrelated—to concurrently mine exclusive and common characteristics among views. A consensus bipartite graph is constructed to extract complementary and consistent multi-view features, leading to superior clustering results. The framework is designed to handle large-scale datasets with over 120,000 samples efficiently due to its linear complexity. Experiments show that NpGC outperforms several classical multi-view clustering methods in terms of clustering accuracy and stability. It is also capable of handling single-view clustering problems and demonstrates strong performance on various real-world datasets. The proposed method is non-parametric, avoiding the need for hyper-parameters, and is computationally efficient. The framework is validated through extensive experiments, showing its effectiveness in capturing high-quality clustering representations. The results indicate that NpGC is a practical and efficient solution for multi-view clustering problems.This paper proposes a non-parametric graph clustering (NpGC) framework for multi-view data, aiming to eliminate hyper-parameters that complicate the clustering process. The framework uses two types of anchors—view-related and view-unrelated—to concurrently mine exclusive and common characteristics among views. A consensus bipartite graph is constructed to extract complementary and consistent multi-view features, leading to superior clustering results. The framework is designed to handle large-scale datasets with over 120,000 samples efficiently due to its linear complexity. Experiments show that NpGC outperforms several classical multi-view clustering methods in terms of clustering accuracy and stability. It is also capable of handling single-view clustering problems and demonstrates strong performance on various real-world datasets. The proposed method is non-parametric, avoiding the need for hyper-parameters, and is computationally efficient. The framework is validated through extensive experiments, showing its effectiveness in capturing high-quality clustering representations. The results indicate that NpGC is a practical and efficient solution for multi-view clustering problems.