
Why we need AI Agents - an LLM can think and an AI Agent can act. an AI Agent wraps an LLM and add features like function calling, memory, human in the loop, calling MCP Servers, atc.
Learn the core concepts of LangChain and how to deploy and configure an LLM in Azure AI Foundry using Terraform.
Build your first ReAct agent step-by-step, connecting to Azure-hosted models and executing reasoning with real prompts.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Discover how AI agents use tools and function calling to extend their capabilities beyond text generation.
Create and bind custom tools to your model, then run a ReAct agent that dynamically selects the right tool at runtime.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Understand the Model Context Protocol (MCP) and how it enables agents to interact with external systems like Microsoft Learn.
Connect your agent to a hosted MCP server and query Microsoft documentation directly through structured tool calls.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Learn how MCP servers can enable real-time web access and extend your agent with external knowledge sources.
Integrate a remote MCP web search server, add error handling, and combine it with a custom webpage-fetching tool.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Explore how to equip AI agents with a secure, sandboxed execution environment using dynamic sessions.
Enable your agent to execute Python and shell commands, upload/download files, and run code safely using Azure Container Apps.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Learn how to add memory to your agents and manage multi-turn conversations using Azure Cosmos DB.
Implement a checkpointer with LangGraph, handle conversation threads, and summarize long interactions.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Understand why human oversight is critical for safe and reliable AI agent behavior.
Pause agent execution for review, then approve, edit, or reject tool actions using Human-in-the-Loop middleware.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Learn how to control and customize agent behavior using middleware hooks before and after model execution.
Implement hooks to inspect, modify, and secure agent interactions at each step of the execution pipeline.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Discover how to design multi-agent systems where specialized agents collaborate to solve complex tasks.
Build an orchestration pattern where a main agent delegates work to sub-agents using MCP-enabled tools.
Github Repository: https://github.com/HoussemDellai/ai-course/tree/main/400_ai_agents_langchain_course
Move beyond a chatbot that only answers questions. Learn how to build AI agents that can use tools, connect to external information, work with conversation context, and involve a human when an action needs review. This course brings those ideas together through LangChain and Microsoft Azure demonstrations.
From language models to connected agents
Start with the foundations of agent-based applications, then extend the same ideas with concrete integrations. You will see how Azure-hosted models, Python code, MCP servers, and Azure services fit into an agent workflow, so the individual features make sense as parts of a working application rather than isolated concepts.
What you will explore
LangChain fundamentals and the ReAct agent pattern: understand the relationship between model reasoning, agent execution, and available tools.
Tools and function calling: define capabilities an agent can invoke instead of relying on text generation alone.
Model Context Protocol (MCP): connect to Microsoft Learn documentation and explore a self-hosted web-search integration.
Dynamic sessions: give agents access to sandboxed Python and shell execution using Azure Container Apps.
Conversation memory: work with context, conversation threads, and Azure Cosmos DB-backed persistence.
Human-in-the-loop workflows: pause execution so a person can review, approve, edit, or reject an action.
Before- and after-model hooks: inspect and customize interactions within the agent workflow.
Sub-agents and orchestration: explore delegation to specialized agents, followed by an introduction to logging and observability with LangSmith.
Learn through implementation
The course pairs short introductions with focused demonstrations. Follow the setup, inspect the code, and see how each capability changes the behavior of the application. The progression helps you understand not just what an agent feature is, but where it belongs and how it connects to the next part of the workflow.
Is this course for you?
This course is for developers and cloud practitioners who want a practical introduction to building agent applications on Azure. Familiarity with Python, basic language-model concepts, and Azure will help you follow the demonstrations. Bring your own project ideas and use these examples as a starting point for designing more useful, connected AI applications.