Previous chapters focused on constructing applications with pre-defined, sequential logic using chains. We now shift to a more dynamic model where a Large Language Model determines the sequence of actions required to accomplish a goal. This is the function of an agent. An agent uses an LLM as its reasoning engine to interact with an environment through a set of available tools.

This chapter introduces the components for building these systems. You will learn about the architecture of agents, tools, and toolkits. We will cover how to supply an agent with both built-in functionalities, such as web search, and custom tools for interacting with specific APIs. You will also examine different agent types and the AgentExecutor that runs the reasoning loop. The chapter culminates in building a practical agent that can find and synthesize information to answer complex queries.