The goal of project MESA (Multidisciplinary Environment for Scientific Advancement) is to build a shared, open-source platform in which scientific data from many fields are automatically described, organized, and connected, so any researcher can find and use them in minutes to hours instead of weeks to months. At its core, metadata-enabled scientific agents read each new dataset, attach descriptions drawn from community-curated standards, and recommend how it can be combined with related information from other disciplines. It will enable researchers spanning diverse disciplines, such as astronomy, biology, environment, public health, and computer science to expedite their discoveries and engage in seamless cross-disciplinary collaboration. The project will be developed and tested with established NSF-supported synthesis centers in environmental data science (ESIIL) and molecular and cellular science (NCEMS), and in AI in agriculture (AIIRA), as well as the international Event Horizon Telescope (EHT) Collaboration. The MESA outreach model ensures that benefits reach real use cases across the range of institutions with diverse levels of research activity and that graduate students, postdoctoral researchers, and professional staff are trained in the design and use of trustworthy, agentic AI for science. MESA will be implemented as a federated, cloud-native data-mesh and data lakehouse coupled to an agentic AI layer operating as an integrated data system and service. There ar