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.

The Particle Accelerator Language Standard (PALS) is a community effort to create an open standard to promote lattice information exchange for particle accelerators. PALS development is a community-wide international effort involving accelerator physicists from multiple institutions. While it started as a lattice standard for beam dynamics simulations, it is now being extended to support other particle accelerator activities, in particular accelerator operation. With new accelerators that are becoming more complex, larger collaborations and the increasing imprint of artificial intelligence in all accelerator activities (from design to operation to workforce development), the imperative for a common, standardized accelerator ontology has been transitioning from “nice-to-have” to “must-have”. We will present the status of the project, its relations to other projects, including to two of the particle accelerator projects of the newly announced US DOE Genesis Mission: the Multi-Office Accelerator Team (MOAT) project and the Nuclear physics AI-Ready Accelerator Data (NARAD) project.

: On November 24, 2025, the U.S. government launched the Genesis Mission (https://genesis.energy.gov/), a national mission to accelerate science through artificial intelligence. As part of Genesis, the US DOE Multi-Office particle Accelerator Team (MOAT) project is a bold and cross-cutting effort to leverage the power of AI at DOE’s current and future particle accelerator facilities. MOAT’s overarching goal is to fundamentally transform how particle accelerators are operated, optimized, and designed by connecting data, expertise, and innovation across the DOE complex. It will ensure DOE’s particle accelerator facilities are at the forefront of AI-enabled scientific infrastructure, and that next-generation accelerators achieve unprecedented performance, efficiency, scientific, and societal impact. We will present the project in the context of the Genesis Mission, its goals, methods and deliverables, and discuss its relationships to other projects and activities in the U.S. and abroad.

The Electron-Ion Collider (EIC) Mission Need requires √s = 20–100 GeV (upgradable to 140 GeV) and luminosity 10³³–10³⁴ cm⁻² s⁻¹. The current ring-ring baseline achieves the full scope, including ~10³⁴ cm⁻² s⁻¹ across all energies. However, when the design is re-optimized for the lower boundary — accepting ~10³³ cm⁻² s⁻¹ and prioritizing cost — an alternative configuration emerges as more advantageous: a hybrid LHeC-like electron accelerator using multi-pass energy recovery linacs (ERL). This solution reduces electron-beam power by roughly an order of magnitude, yielding nearly a factor of two reduction in total project cost compared with the present baseline while still satisfying the minimum physics requirements. The study performs parametric cost and performance modeling, augmented by AI-driven optimization, to explore this design space. Serving primarily as an educational exercise for the next generation of accelerator physicists and engineers, the paper demonstrates modern design methods: rapid parametric scans, cost-driven optimization, and integration of AI tools. It examines technical feasibility, identifies critical R&D (high-current ERL operation, beam–beam effects, synchronization, etc.), and discusses how such re-optimization studies can be used to train designers in an era when artificial intelligence dramatically expands exploration of complex accelerator parameter spaces.