Large Language Models (LLMs) are helping developers build innovative applications due to their ability to process and produce nuanced language. However, LLMs have limitations as they are not connected to the external world and lack access to real-time information or the ability to take actions like booking flights or sending messages. To make LLMs more useful, they need to interact with the outside world by calling APIs. For example, to build a chatbot that provides real-time weather updates, developers need to write a prompt that instructs the LLM to call a weather API when necessary. The application code then executes the API call and feeds the response back to the LLM to produce a user-friendly response. The process involves designing a prompt that specifies when an API call is needed, writing the application code to execute the API call, and feeding the API output back to the LLM. The prompt should include an instruction that tells the LLM what the objective is and describe the tools it can use. An example input and response can help the LLM understand the behavior required. The application code detects when the model outputs a tool command, executes the API call, and processes the response. The response is then rephrased into a conversational format using the LLM. This method can be extended to multiple tools by listing out all the tool commands and using more sophisticated prompts, such as the ReAct technique, which combines reasoning steps with actions. Prompting methods are an active area of research, and developers are encouraged to explore different techniques to teach LLMs how to use external tools.