Although AI agents (with increasingly-large language models) have become central to applications in healthcare, finance, education, and customer service, there is currently no bibliometric study that consolidates and evaluates this rapidly expanding body of literature in a systematic way. The purpose of this study is to give the first comprehensive bibliometric analysis of scholarly research on AI agents, including mapping the intellectual, conceptual, and social structure of the field in order to better understand the evolution of the field and emerging trends. A specific query of the Scopus database was used to extract 441 articles published between 2020 and early 2026. Data was cleaned and harmonized using OpenRefine and BiblioMagika and analyzed using Bibliometrix (R package) and VOSviewer. The analyses comprised descriptive statistics, citation metrics (h-index, g-index, m-index), trend topic analysis, cloud words and international collaboration mapping. Results indicate that publications have a high growth rate since the year 2020, with China, the United States and Europe being the major contributors. Regarding the most influential, the most authoritative influencers are ZHANG Y and CHEN Y; the trend topics analysis indicates the thematic development from artificial intelligence, language models, and artificial intelligence agents. Europe, Asia, and Australia have evident cooperation, but not Africa and South America. This study has theoretical and practitioner implications. Theoretically, this would make the artificial intelligence agents the separate and high-speed research streams of artificial intelligence. Practically, it offers some insightful results for researchers, institutions, and policy makers, for example, to identify major contributors, patterns of international collaborations, and trends for applied research.