Speaker
Dr
Andrew Owen
(Center for High Throughput Computing)
Description
The interface between researchers and computing systems is shifting: users are reading documentation even less than before and are instead asking AI agents for answers. That means in order to get users the correct, up-to-date information, we have to make sure the AI agent can as well. We've developed a prototype tool "CHTC Help!" based on the Model Context Protocol (MCP) to bridge the gap between AI agents and local system information. We'll discuss the "why" and "how" of our implementation, the pros and cons we've discovered, and how you can adapt the tool to your own system.