Large language models (LLMs) are increasingly applied in biomedical research for tasks such as data annotation, literature analysis, and knowledge retrieval, but their usefulness is often limited by hallucinations, missing domain expertise, and restricted access to specialized resources. While recent agent-based LLM systems can use external tools to overcome some of these limitations, most existing solutions remain narrowly tailored and difficult to reuse or extend. We introduce BioContextAI, an open-source initiative for building modular, agent-based biomedical research assistants from standardized components. BioContextAI enables seamless integration of diverse resources, including biomedical knowledgebases, medical literature, ’omics data, and specialized analysis tools that are often inaccessible to conventional web-based approaches. By emphasizing interoperability and extensibility, the framework supports flexible and reliable AI-driven research workflows. Importantly, BioContextAI is designed in line with FAIR principles for research software, promoting reuse and long-term sustainability.