Course Objective & Scope
Simple agent loops that look promising in a prototype often fail in production due to lost state, runaway loops, weak type safety and unvalidated tool calls. This training teaches how to engineer enterprise-grade single- and multi-agent systems using the Microsoft Agent Framework, which unifies the strengths of Semantic Kernel and AutoGen.
Unified agent abstractions in Python and .NET, function and Model Context Protocol (MCP) tools, combining deterministic process steps with autonomous reasoning, multi-agent handoff and group collaboration patterns, and OpenTelemetry observability are covered hands-on. Participants build a reusable multi-agent workflow and architecture checklist tailored to an internal business scenario.
Who Should Attend?
- Software engineers building enterprise AI applications in Python or C#/.NET
- AI engineers designing multi-step autonomous workflows and assistants
- Solution architects integrating ERP, CRM and internal APIs with agent systems
- Prerequisite: Python or C# programming and familiarity with LLM APIs and tool calling
Key Takeaways
- Build model-agnostic, type-safe agents and plugins using the Microsoft Agent Framework
- Connect enterprise APIs and MCP servers with strict input validation and authentication filters
- Select and implement sequential, concurrent, handoff and group-chat multi-agent topologies
- Bound agent autonomy using human-in-the-loop approvals, state checkpointing and step/token budgets
- Verify agent behaviour before release using OpenTelemetry traces and structured evaluation suites
Curriculum
Microsoft Agent Framework Architecture: Unifying Semantic Kernel and AutoGen
- Core agent abstractions, message types, thread/state management and Python/.NET parity
- Provider-agnostic execution across Azure AI Foundry, OpenAI and open-weight models
Tools, Plugins and Model Context Protocol (MCP) Integration
- Type-safe function schemas, OpenAPI plugins and MCP client connectivity
- Pre/post tool invocation filters, authorization checks and graceful error recovery
Multi-Agent Orchestration Patterns
- Handoff, Sequential, Concurrent and Magentic/GroupChat coordination patterns
- Termination conditions, conversation summarization and context-window budgeting
Stateful Processes and Human-in-the-Loop Controls
- Blending deterministic business rules with autonomous LLM steps in event-driven workflows
- State checkpointing for long-running processes and human approval gates for high-impact actions
Observability, Security and Enterprise Deployment
- Tracing every agent turn, tool invocation and token spend with OpenTelemetry
- Agent-to-Agent (A2A) interoperability and deploying to Azure Container Apps / Foundry Agent Service
Other Courses in This Category
-
Accelerating the software development lifecycle with GitHub Copilot Agent Mode, Copilot Coding Agent, custom instructions, IDE MCP servers and automated PR review.
- Live Course
- 5 modules
-
Deploying agents built with any framework (Strands, LangGraph, CrewAI) securely and at scale using Amazon Bedrock AgentCore: Runtime, Memory, Identity, Gateway and Observability.
- Live Course
- 5 modules
-
Building model-driven agents with the open-source AWS Strands Agents SDK: @tool decorators, native MCP integration, multi-provider model support, multi-agent patterns and OpenTelemetry tracing.
- Live Course
- 5 modules
-
Managed enterprise RAG with Amazon Bedrock Knowledge Bases (advanced parsing, chunking, hybrid search, reranking) and safety enforcement with Bedrock Guardrails (PII, grounding, content filters).
- Live Course
- 5 modules
-
Training, deploying and governing classical ML models and open-weight LLMs on AWS using Amazon SageMaker AI: Unified Studio, JumpStart, Training, Pipelines, Model Registry and Endpoints.
- Live Course
- 5 modules
-
Enterprise AI and agent development on Google Cloud Vertex AI with Gemini models: Google Gen AI SDK, grounding with Vertex AI Search, context caching, safety controls and deployment.
- Live Course
- 5 modules
Get in touch
