A tool call is a model-generated request to invoke an external function through an application layer. Instead of producing only text, the model outputs a structured action such as “search,” “query,” “scan,” or “send,” and the surrounding software decides whether to execute it. Tool calls are the bridge between reasoning and real-world actions in agentic systems.
In cyber security, tool calls matter because they can turn a chatbot or AI agent into a control interface for sensitive operations. If access control is weak, an attacker may use a chat message to trigger reconnaissance, file access, network requests, or other privileged steps. Defenders should treat tool execution as part of the attack surface: require approval for risky actions, restrict what tools are exposed, log every call, and limit outbound network access. Good security depends less on prompt wording and more on what the application permits the model to do.