Orchestrate with Agents & MCP: Building Advanced Agentic Systems
Description
Large Language Models (LLMs) are now a well-established concept, and AI agents are gaining traction in the market. But how do we assemble them into a cohesive, orchestrated system that performs truly complex processes?
This workshop takes agentic thinking to the next level. We will explore why a single agent is often insufficient and delve into building collections of agents that collaborate seamlessly. Crucially, we'll understand when an "orchestrator" is needed to conduct this agentic symphony. You'll learn how to effectively integrate tools, APIs, business logic, and memory – cutting through the hype with a practical approach that bridges concepts to real-world implementation.
As a grand finale, we'll deep-dive into MCP (Model Communication Protocol): a novel protocol enabling models to comprehend complex contexts, access tools, files, and systems, transforming intelligent agents into fully functional components within real-world systems.
This is not another LLM workshop. This is the transition from "I ask, it answers" to "I plan, they act."
Intended audience
AI Practitioners, Developers, Programmers, Software Engineers, Data Scientists, Solution Architects, and Tech Leaders
Topics
Introduction to Agentic Systems
Solving problems with a team of Agents.
Level setting: LLMs revisited.
Defining what an Agent is and how it differs from a simple chatbot.
Defining and utilizing "Tools" within agentic contexts.
Agentic workflow patterns: roles, boundaries, and task decomposition.
Mapping a complex problem onto a team of agents.
Practical Exercise: Agent specification based on a given challenge.
Building the Foundation for Smart Agents: Advanced LLM Techniques
Chain of Thought: When it makes all the difference in results.
Reasoning techniques for agents.
Grounding: Retrieval Augmented Generation (RAG) in depth.
Practical examples of advanced LLM techniques.
Practical Exercise: Developing a simple RAG agent.
LangChain: The Agent Assembly Kit
Introducing LangChain: A comprehensive framework for agent development.
Core Entities: Indexes, Chains, Prompts, Agents.
Integrating memory and tools into LangChain agents.
Case Study: A conversational agent that thinks, initiates, and acts.
Practical Exercise: Building a conversational agent in LangChain with integrated tools.
Teamwork: Multi-Agent Systems
Introduction: Why multi-agent systems are essential.
Introducing AutoGen: A framework for multi-agent conversation.
Scenario: A "digital team meeting" simulation.
Guided Exercise: Specifying and Building a Multi-Agent Team with AutoGen
Task specification for a multi-agent solution.
Team definition and role assignments for agents.
The "Committee Chairperson" role in orchestration.
Establishing a "License to Operate" for agents.
MCP: Empowering Agents to Be Superheroes
Introduction to MCP (Model Communication Protocol).
MCP as "Extension Packs" for agents.
Understanding MCP Servers.
Understanding MCP Clients.
Real-world scenario demonstration with MCP.
MCP: Summary and Future Outlook
Where is this technology heading?
What is the ultimate role of humans in agentic systems?
Inspiration for continued exploration.
Q&A and Feedback.
Summary
האם אתה בטוח שאתה רוצה לסגור את הטופס ולאבד את כל השינויים?
