Ocean basins are fundamental units for modeling the Earth system, paleoceanography, and the global carbon cycle. However, their boundaries are often defined heuristically, limiting the robustness of reduced-order models and the interpretation of paleoproxy data, especially in data-limited paleo- or planetary contexts. We present ExoCcycle, an open-source Python library for objective, automated community detection on spherical grids. This framework implements novel composite algorithms (e.g., SB-Reduction) that couple efficient partitioning (Leiden/Louvain) with ensemble-based agglomerative clustering for robust boundary detection. A key technical innovation is our Difference Quantile Transformation Cumulative Density Function (DQT-CDF) edge-weighting scheme, enabling the principled analysis of single or multiple, non-normally distributed scalar fields in a large spherical domain. We validate the method using modern bathymetry and temperature/salinity fields, demonstrating that (1) a spatial resolution of 1–2 degrees is necessary to capture critical basin-defining features such as ridges and plateaus; (2) basin boundaries evolve significantly over geological time, underscoring the inadequacy of using static, modern boundaries for past climate simulations; and (3) the ocean's community structure is fundamentally layered – deep basins (defined by bathymetry) are distinct from shallow shelf partitions (shaped by sedimentation, sea-level changes, and riverine fluxes), and surface basins (driven by wind and temperature/precipitation). ExoCcycle provides a systematic tool for generating physically-grounded, time-evolving basin definitions, enabling the development of next-generation modular intermediate-complexity models for Earth and exoplanet habitability. As a generalized spherical community detection tool, our new framework is also broadly applicable to other non-ocean related domains, including ecology and land processes, atmospheric science, solid-Earth geophysics, and planetary science.