This chapter introduces the core concepts and setup needed to build agentic workflows using LlamaIndex. Get hands-on practice with tools like Tavily and OpenAI models, define your first tool function, and create an intelligent agent capable of interacting with the web. By the end, you'll have a fully functional AI agent built with LlamaIndex, ready to tackle real-world tasks.

Welcome!

Getting started with LlamaIndex

What are agents?

Creating an agent

Do agents remember?

Maintaining state across agent runs

Streaming output and events

Building an Agent

Why stop at one agent when you can build a team? In this chapter, you'll unlock the power of multi-agent systems, where specialized agents work together like a research squad, each with a unique role. You'll build your own collaborative agent workflow that searches, writes, and reviews, showing how teamwork makes your AI smarter and more scalable.

What are multi-agent systems?

Multi-agent system with AgentWorkflow

Building agentic workflows from scratch

Designing multi-step workflows with custom events

Concurrency and event collection

Putting together a multi-agent system

Adding self-reflection to the workflow

Wrap-up

Creating a Deep Research Workflow

Ready to build your own team of AI agents? In this course, you’ll start with a single smart assistant and level up to full agentic workflows that plan, search, remember, and collaborate. Using LlamaIndex, you’ll create powerful systems that tackle complex tasks like deep research and reporting. Along the way, you’ll learn how to orchestrate tools, memory, and multi-agent collaboration to bring your AI workflows to life.

<h2>Build Smart, Capable AI Agents from Scratch</h2>

Learn how to build intelligent agentic workflows from the ground up using LlamaIndex, a powerful framework for orchestrating AI behavior. Starting with a single-agent setup and moving toward multi-agent collaboration, you'll create a system that can perform research, generate reports, and reflect on its own outputs.
<br></br>

<h2>Create and Orchestrate Agentic Workflows</h2>

You’ll begin by setting up a simple AI agent using the AgentWorkflow class and connecting it to functionalities such as searching the web. You’ll learn how to give your agents “memory” using context and make their responses feel more dynamic with streaming outputs. As the course progresses, you’ll explore advanced capabilities like event handling, concurrent execution, and creating custom workflows using step-based logic. You'll also implement self-reflection loops to allow agents to improve their own outputs.
<br></br>

<h2>Design Multi-Agent Systems with LlamaIndex</h2>

In the final step, you’ll move beyond single-agent systems by designing teams of AI agents that work together. You’ll build workflows where each agent takes on a specialized role—like researching, writing, and reviewing—and shares information through a common context. You’ll experiment with sequential, looping, and branching workflows, gaining insights into how to manage complex tasks with multiple agents.

Intermediate Python

Build AI agentic workflows that can plan, search, remember, and collaborate, using LlamaIndex.

Adding self-reflection to the workflow

Building Agentic Workflows with LlamaIndex

Anleitung zur Übung

Interaktive Übung