Agentic AI – LangGraph and MCP on Azure

Candidates for this training should have a foundational understanding of generative AI (how LLMs work, token concepts, and the difference between traditional AI and generative AI). Practical knowledge of Python is also required.

Course materials and hands-on labs are provided in English. A B1 level of English is recommended. You can find more information on language proficiency levels via the following link: Language proficiency level classification.

Understanding agentic AI and why it’s the “next step”

• Situating agentic AI in the continuity of generative AI: from chatbots to autonomous agents.
• Understand the limits of the classical conversational approach and what the agentic brings.
• Discover the fundamental concepts: reasoning, planning, use of tools, agentic loop.
• Explore the ReAct (Reasoning + Acting) pattern and how it works.
• Ecosystem overview: LangGraph, Semantic Kernel, AutoGen, CrewAI.

Discover LangGraph and build a first agent

• Understand the architecture of LangGraph: graphs, nodes, states, and transitions.
• Learn about the role of Azure OpenAI as an agent reasoning engine.
• Build a first LangGraph agent with Azure OpenAI capable of using simple tools (search, calculation).

Enriching your agent: memory and robustness

• Add conversational memory and checkpointing to an agent.
• Understand the human3in3the3loop principle for sensitive actions.
• Manage errors and edge cases in the agentic loop.
• Enrich the agent of the previous lab with memory management and human validation on certain actions.

Understanding the Model Context Protocol (MCP)

• Discover MCP: origins, objectives and positioning in the agent ecosystem.
• Understand the MCP architecture: Host, Client, Server and the primitives (tools, resources, prompts).
• Identify use cases: Connect an agent to data sources, APIs, or documentation.
• Check out the Microsoft Learn MCP server as a real-world example.

Connect your agent to an MCP server

• Connect the LangGraph agent to the MCP server in the Microsoft Learn documentation. The agent becomes able to search and synthesize technical documentation to answer questions.
• Test, iterate, and refine agent behavior when faced with different types of queries.

Integrate a Gradio interface and deploy on Azure

• Learn about Gradio and its integration with AI agents.
• Understand the deployment architecture provided: FastAPI server, Gradio, Docker containerization, Azure Container Apps.
• From a provided repository (FastAPI server, Dockerfile, CI/CD pipeline, Terraform), integrate the agent developed during the training to the Gradio interface, customize the interface by following the Gradio documentation, and then deploy the whole thing to Azure.

This training combines theory with hands-on workshops and technical demonstrations to ensure participants become quickly operational. Each participant also receives course materials.

The training is delivered by one of our consultant-trainers. With strong real-world experience, they make the learning experience both interactive and enriching.

For assessment, the trainer uses regular questions and various methods to continuously evaluate learning progress. This approach ensures a dynamic and engaging learning experience.

After the training, participants are asked to complete a satisfaction questionnaire. Your feedback helps us continuously maintain and improve the quality of our courses. In addition, for proper tracking, each participant signs an attendance sheet for each half-day session.

Finally, we offer flexibility: the training can be delivered either in-person or remotely, and it can be customized to meet your company’s specific needs upon request.

You can register for one of our training courses up to five business days before the start date, provided that places are still available and we have received your signed quotation.

If you have specific needs related to a disability, please let us know. We are happy to adapt our delivery methods according to the type of disability.

Our training center, Cellenza Training, is located at 892 Rue Yves Kermen, 92100 Boulogne-Billancourt.

You can reach us easily by public transport:

• Take Metro Line 9 and get off at Pont de Sèvres
• Take Metro Line 10 to Boulogne Jean Jaurès
• Or take Tram T2 to Brimborion