Agents / Standard term
Tool use / function calling
The capability that lets an AI model take real actions: the model outputs a structured request ("call this function with these arguments"), and your application decides whether to run it.
Tool use turns a text generator into something that can take real actions. When a model needs to do more than generate text, it outputs a structured request: a function name and its arguments. For example: {"function": "send_email", "to": "sam@example.com", "subject": "Meeting Tuesday"}. Your application receives that request, validates it, runs the code, and sends the result back. The model uses that result to continue its work. This is how chatbots become agents: tool use lets AI book a meeting, look up inventory, edit a file, or send an email.
Builder example
Every real-world action your AI product takes flows through a tool call, making this the most important place for guardrails. Validate the arguments before running any function (the model can hallucinate parameters). Limit which tools each workflow can access (a summarization task should never send emails). Log every tool call separately from the conversation so you can audit what actually happened.
The user says, 'Book Sarah for Tuesday afternoon.' The model proposes a calendar-create call with attendee, time, title, and timezone.
Validate the fields, confirm ambiguity, and require approval before sending invites.
Common confusion: The model never executes code itself. It produces a structured request that says "I want to call this function with these arguments," and your application code controls whether and how to proceed.