AI agents often require access to external information or functionalities not directly available within the language model. To perform actions such as searching the web, executing computations, or interacting with databases, agents must be equipped to use tools. This chapter examines the prompt engineering techniques that enable AI agents to select, operate, and manage these external tools.

Specifically, you will gain understanding of:

• The fundamental concepts behind language models augmented with external tools.
• Prompting techniques for guiding an agent's tool selection and correct operation.
• Methods for formatting tool descriptions and API specifications to be comprehensible to LLMs.
• How to use prompts to manage the inputs to tools and the outputs received from them.
• Approaches for prompting agents to interact with APIs and process their responses.
• Prompt design for helping agents handle errors that occur during tool execution.
• Strategies for creating prompts to integrate novel or custom-developed tools within agentic systems.

Mastering these techniques will allow you to design prompts that significantly extend your agents' capabilities by connecting them to a wider range of resources and functions.