Operating large-scale scientific facilities requires coordinating diverse subsystems, translating operator intent into precise hardware actions, and maintaining strict safety oversight. Language-model agents offer a natural interface for these tasks, but most existing approaches are not yet reliable or safe enough for production use. We introduce Osprey, a framework that wraps a coding agent in a control-room operator interface, a tool surface that reaches hardware through pluggable connectors for the control system used in our community, and a first-class component for natural-language search of facility electronic logbooks. The agent itself is treated as a replaceable component: operator interface, safety policy, tool servers, and connectors stay under facility control, while the agent backend can be swapped as the AI ecosystem evolves. A declarative build-profile mechanism lets each facility maintain its own configuration without forking the shared framework, keeping deployments reproducible across updates. Osprey has been deployed at several DOE accelerator facilities through the MOAT seed effort within the Genesis~Mission. This paper presents the current framework architecture and reports on the substantial evolution Osprey has undergone over the past year.

Agentic artificial intelligence (AI) systems are emerging as a practical middle layer for intelligent, self-optimizing accelerator operations. Building on work at the Advanced Light Source, we have developed a modular agentic framework that integrates natural-language interfaces with control systems, archival data, simulation tools, and technical documentation, enabling context-aware reasoning with human-in-the-loop execution. This approach provides intuitive access to complex accelerator environments while preserving safety, transparency, and reproducibility. A central focus of the framework is usability and rapid onboarding. Self-contained tutorials, reproducible deployment patterns, and a facility-agnostic interface allow laboratories to adopt agentic workflows with minimal customization. This streamlined process has supported deployments at APS, SLAC, SNS, CEBAF, ALS, and BELLA as part of a DOE/MOAT effort within the Genesis mission, where agents execute multi-step tasks, generate inspectable plans, and analyze historical and live data through a shared language interface. This contribution presents the core architecture, cross-facility deployment experience, and early operational lessons from these implementations. It also outlines how agentic workflows form a unifying layer for emerging capabilities, such as physics-informed optimizers, reinforcement-learning agents, and automated tuning assistants, advancing autonomous control in next-generation scientific facilities.