Design multi-agent architectures where specialized sub-agents collaborate on complex tasks. You'll learn to implement researcher/editor and other collaboration patterns, manage inter-agent communication, and orchestrate workflows where agents with different capabilities work together; the key pattern for building enterprise-scale agent systems. This is the most architecturally complex short course in the program — multi-agent systems are the pattern for handling task complexity that overwhelms a single agent.

Embed the Copilot SDK into an existing application with event-driven triggers, domain-specific tools, and user-facing interaction surfaces. You'll learn the integration patterns that transform standalone agent scripts into embedded application features, moving from "agent as a tool" to "agent as a product capability."

Design domain-specific agent workflows and validate them through integration testing. You'll learn to define end-to-end agent workflows for specific business domains, write integration tests that verify tool invocations, session behavior, and output quality, and build confidence that embedded agents behave correctly before production deployment.

Explore how generative AI tools can accelerate multi-agent system design, workflow specification, and integration test generation. You'll learn to use AI as a design accelerator for complex architectural tasks while critically evaluating AI-generated architectures and test suites.

Build an embedded multi-agent application that combines model routing, multi-agent orchestration, application integration, and integration testing into a comprehensive portfolio artifact. This is the most complex project in the program, combining multi-model routing, multi-agent orchestration, application embedding, and integration testing into one cohesive deliverable.