Agent concepts
Tool call
A tool call is the moment an AI agent invokes an external function, with concrete arguments, to take an action or fetch data.
Also called: function call, tool use
Definition
What is a tool call in AI agents?
A tool call is the unit of action for an AI agent. The model emits a structured request naming a tool and the arguments to pass it, a runtime executes that function, and the result is fed back into the model. Reading a file, running a query, sending a message, and charging a card are all tool calls.
The tool call is the right place to enforce security, because it is where an agent's intent becomes a concrete, inspectable action. The tool name and its actual arguments reveal what is about to happen: which bucket is about to be deleted, which secret is about to be sent, where data is about to go. Scoring and policing the call before it runs stops a bad action while it is still just a request.
FAQ
Common questions.
Why inspect the tool call instead of the prompt?
The prompt shows intent in words; the tool call shows intent as a concrete action with real arguments. Inspecting the call catches the actual dangerous operation, including cases where a benign-looking prompt produces a harmful call.
Related terms
Keep reading.
Govern the actions, not just the vocabulary
AxioRank scores every tool call your agents make for leaked secrets, PII, destructive operations, and prompt injection, checks it against your policy, and proves it in a tamper-evident audit log. Start free, no card.