Researchers have developed Denario, an AI-powered scientific assistant designed to accelerate the research process by helping scientists generate hypotheses, analyze data, and write manuscripts. Created by a team from the University of Cambridge, the Flatiron Institute, and the Autonomous University of Barcelona, Denario uses large language models to support every stage of scientific inquiry, from idea generation to final documentation. Unlike current tools that handle only isolated tasks—like writing abstracts or visualizing data—Denario integrates multiple AI agents into a modular, end-to-end system capable of synthesizing literature, formulating new research questions, analyzing datasets, and producing draft papers. Denario’s architecture is built around specialized AI agents, each responsible for a distinct function. When a scientist uploads a dataset and a brief goal, the system begins by generating and refining research ideas. A second set of agents then scours existing scientific literature to ensure the proposed project is novel and grounded in prior work. Methods and planning agents suggest analytical approaches, which are executed by CMBAgent, a multi-agent system that writes, debugs, and runs code, then interprets results. Finally, writing and reviewing agents draft and refine summaries of findings. The modular design allows users to employ individual components for specific tasks, such as data analysis or literature review, or use the full pipeline for comprehensive support. In tests across astrophysics, neuroscience, chemistry, biology, and materials science, Denario was used hundreds of times. While most outputs were deemed unhelpful, about 10% produced interesting or novel research insights. The system’s interdisciplinary nature is a key strength—by drawing from diverse fields, it can suggest connections a specialist might overlook, potentially sparking innovative research directions. Researchers hope Denario will help scientists reclaim time from repetitive tasks like searching databases, formatting figures, or summarizing results, allowing more focus on creative, high-level thinking. Despite its promise, Denario has significant limitations. It occasionally fabricates data or misrepresents findings—so-called “hallucinations”—and has struggled to clearly reference prior studies or convey uncertainty in results. The team had to explicitly instruct the system not to generate “dummy data” after it produced false information. Human oversight remains essential, and the current version is not a replacement for scientific expertise. Ethical and technical challenges remain, including concerns about copyright, authorship, and the reliability of AI-generated content. The team emphasizes that Denario is a tool to augment, not replace, human researchers. They are working on future versions to improve efficiency, filter low-quality outputs, and better handle scientific nuance. Denario’s development reflects a growing effort to apply AI to the full research lifecycle, made possible by advances in large language models like ChatGPT, Google Gemini, and Anthropic’s Claude. The project was made possible through collaboration across disciplines, including physics, biology, machine learning, and philosophy. The researchers call for an open discussion on how to responsibly integrate such tools into science, ensuring they enhance discovery while maintaining integrity. With its potential to speed up research and foster cross-disciplinary innovation, Denario represents a significant step toward a new era of AI-augmented science.