Osprey: A Scalable Framework for the Orchestration of Agentic Systems

Coordinating workflows across complex systems remains a central challenge in safety-critical environments such as scientific facilities. Language-model-driven agents offer a natural interface for these tasks, but existing approaches often lack scalability, reliability, and human oversight. We introduce the Osprey Framework, a domain-agnostic, production-ready architecture for scalable agentic systems that integrate conversational context with robust tool orchestration across safety-critical domains. Our framework provides: (i) dynamic capability classification to select only relevant tools; (ii) plan-first orchestration with explicit dependencies and optional human approval; (iii) context-aware task extraction that combines dialogue history with external memory and domain resources; and (iv) production-ready execution with checkpointing, artifact management, and modular deployment. We demonstrate its versatility through two case studies: a deployment at the Advanced Light Source particle accelerator and a tutorial-style wind farm monitoring example. These results establish Osprey as a reliable and transparent framework for agentic systems across diverse high-stakes domains.