AI is moving beyond single models towards autonomous agents that can reason, act and collaborate across tools and systems. As this shift accelerates, a critical challenge emerges: how do agents reliably access, share and act on context in complex, real-world environments?

This session examines how Foundry-based AI platforms, combined with the Model Context Protocol (MCP), enable teams to move from experimentation to reliable, scalable agent systems. Focusing on real architectures and deployed use cases, we explore how governance, security and operational control are achieved in practice, how MCP standardises agent interaction with enterprise systems, and what trade-offs matter when building agents that must perform under real-world constraints.