Microsoft 365 Agents SDK and Copilot Extensions

• Solution architects and AI engineers
• Microsoft 365 developers extending Copilot
• Technology consultants designing enterprise AI integrations
• Platform engineers enabling secure AI adoption

• C# basics at the level of the GOC2125 course
• JavaScript basics at the JS_PROG1 course level
• TypeScript basics at the JS_TS1 course level
• Knowledge of REST & JSON
• Azure & Microsoft 365 basics
• Previous work with Microsoft Graph (optional/advantageous)
• Experience with prompt engineering (optional/advantageous)

1. Copilot architecture & Agents SDK overview (2 hours)

• Extension options: Agents SDK, plugins, Graph connectors, actions
• Azure Bot Service channels and adapters; activities and "turns"

2. Microsoft 365 Agents Toolkit & Playground (2.5 hours)

• Templates for Visual Studio/VS Code, scaffolding, local Agents Playground
• Supported languages (C#, JS/Node, Python) and prerequisites

3. Responsible AI & governance (2.5 hours)

• Guardrails, safety filters, content moderation
• Privacy, compliance & traceability

4. Corporate ingestion strategy (2.5 hours)

• Principles of SharePoint/OneDrive content indexing
• Graph connectors & metadata enrichment

5. Vectorization & RAG design (2.5 hours)

• Chunking, embeddings, hybrid semantic ranking
• Latency vs. relevance trade-offs

6. Prompt & system orchestration (2.5 hours)

• System vs. user vs. tool messages
• Few-shot, prompt chaining, evaluation heuristics

7. Application agent, manifest & API surface (2.5 hours)

• AgentApplication/AgentApplicationOptions basics; endpoint mapping
• Manifest, capability declarations, versioning

8. Action handlers & integration with Microsoft Graph (2.5 hours)

• Action routing
• Parameter validation
• Permission boundaries

9. Middleware & enforcement policies (2.5 hours)

• Pre/post processing, editing, output filtering
• Hallucination mitigation patterns

10. Event-driven orchestration (2.5 hours)

• Azure Functions vs Logic Apps vs Durable orchestrations
• Model and tool call chaining

11. Evaluation & quality (2.5 hours)

• Qualitative vs quantitative metrics, human-in-the-loop
• Golden set & regression tests

12. Performance, costs & scaling (2.5 hours)

• Token economy, cache layers, retrieval optimization
• Resilience and timeout patterns

13. Security & risk management (2.5 hours)

• Exfiltration prevention, scope minimization
• Output sanitization and watermarking (where appropriate)

14. Deployment & lifecycle management (2.5 hours)

• Propagation between environments, versioning, rollback
• Model exchange governance & deprecation planning

15. Capstone architecture & roadmap (2–2.5 hours)

• Presentation of end-to-end solution with Agents SDK + RAG
• KPI review & evaluation matrix
• Next steps: AI-102, AZ-204, multi-agent patterns

Previous courses

Programming with C# Language II [GOC2125]

JS_PROG1 - JavaScript I: Basics [JS_PROG1]

TypeScript - type checker for JavaScript [JS_TS1]