Session Abstract

This session explores advanced design patterns for building resilient multi-agent systems in Microsoft Copilot Studio, focusing on nested (child) agents, connected agents, Agent-to-Agent (A2A) collaboration, and MCP-enabled autonomous agents. Participants will learn how to model roles, capabilities, and boundaries across agents, implement planner/executor and supervisor/child topologies, and apply guardrails for safe scaling. Key topics include: Nested Agents: Lifecycle management, context scoping, memory sharing, and escalation strategies. Connected Agents: Designing stable contracts for plugins/connectors and managing capability registries. A2A Collaboration: Message routing, negotiation, and observability practices for reliable handoffs. MCP Integration: Leveraging Model Context Protocol for capability discovery, permissioning, and safe autonomy. Governance & Safety: Least-privilege design, auditability, and rollout strategies to minimize risk. By the end of the session, attendees will be equipped to design robust multi-agent architectures, enable scalable collaboration, integrate MCP tools, and apply testing and observability practices to improve outcomes. Audience: Solution architects, developers, and technical leads working on Copilot Studio or multi-agent systems. Level: Intermediate with advanced design patterns.

Holger Imbery

As a Principal Architect and Microsoft MVP, I am leading the design and deployment of agentic AI systems that enable autonomous decision-making and intelligent process automation across enterprise environments. My passion lies in helping organizations do more with less - leveraging Microsoft technologies to transform business operations and customer experiences.