Stream: Wellington A
Time: 10:00 - 10:45
This session walks through how we built a governed Model Context Protocol (MCP) layer that lets AI agents query and update z/OS data sources in natural language, and how QA, support and development teams use it day to day. We start with the problem: accessing ADABAS, Db2 for z/OS and IMS each requires specific subsystem knowledge, query languages and access paths — skills concentrated in a shrinking pool of experts. QA and support are routinely blocked waiting on those experts just to confirm a record, read log data, or check a table. We then show the architecture and demonstrate it live against three data sources: - ADABAS — retrieving business data. - Db2 for z/OS — turning plain-language questions into SELECTs, and reading Db2 log data. - IMS — reading and writing segment data (insert, update, delete), using the IMS catalog for segment and DBD metadata, scoped by PSB. Attendees will learn what MCP is and why an AI-protocol standard fits this problem, and how the same MCP layer supports three distinct workflows for developers, QA engineers and support analysts. Attendees leave with a concrete, reusable architecture pattern for giving non-specialists safe access to mainframe data.
There is currently no attachment for Closing the Skills Gap with AI: Natural-Language Access to Mainframe Data via MCP
Miguel Fernandez is a Senior Member of Technical Staff at Salesforce, with nearly 30 years working at the core of enterprise data systems. He spent over two decades at Informatica developing log-based Change Data Capture and bulk data access for IMS, ADABAS and Datacom, and more recently modernised build-and-test pipelines across z/OS, Windows and Linux using Jenkins and Zowe. His focus is making mainframe data accessible to the teams who depend on it.
Click here to give some Feedback so we can make it even better next year!