#VoxxedDaysCERN26

At CERN’s Large Hadron Collider (LHC), every minute of downtime is precious. During technical stops, hundreds of specialists perform maintenance and upgrades before physics runs resume. Coordinating these interventions requires precise, conflict-free scheduling - a complex and time-critical process.

During the AI for Scientific Writing and Coding Assistants Hackathon, we built a prototype Model Context Protocol (MCP) server to assist LHC coordinators in planning and analyzing linear schedules. The system bridges the LHC scheduling tools’ APIs with a CERN-hosted large language model, enabling coordinators to interact with complex planning data in plain English. Through this interface, the MCP can perform automatic analyses such as detecting potential conflicts, validating resource and timing constraints, and gathering contextual information about interventions.

By combining structured scheduling data with natural language understanding, the prototype demonstrates how AI-assisted tools can enhance situational awareness and streamline planning quality checks. This talk explores the motivation, design, and lessons learned from integrating MCP with existing LHC workflows - showcasing how lightweight AI infrastructure can empower experts to coordinate maintenance and upgrades in one of the world’s most complex scientific machines.

Devoxx

A CERN developer presents how they built an MCP server to help coordinate scheduling at the Large Hadron Collider. The LHC scheduling problem involves thousands of tasks with complex physical, safety, and technical constraints, all managed in a massive MS Project file. The solution exposes the scheduling backend's APIs via an MCP server so coordinators can query conflicts in plain English using a local LLM. Key lessons: keep business logic in code rather than delegating it to the LLM, shape data to minimize context window usage, and return clear structured signals (e.g. a conflict list) rather than raw data dumps for the LLM to reason over.

MCP at the LHC: Coordinating the World’s Largest Machine by Georg Šumailov