Glossary
Tool Calling
Also known as: function calling, tool use (LLM)
Definition
Tool calling, sometimes called function calling, is the capability that lets a large language model invoke external functions, APIs, or services as part of generating a response. The model emits a structured request specifying which tool to call and with what arguments; the runtime executes the call and feeds the result back into the model's context.
Why it matters
Tool calling is what turns an LLM from a text generator into an agent. It is the mechanism behind every production AI workflow that fetches a credit score, looks up a policy, calls a payment API, or schedules a meeting. It is also the highest-leverage failure surface: a wrong tool, wrong argument, or successful prompt injection that triggers an unintended call can have real-world consequences (a denial issued, a payment sent, a record exposed).
For regulated workloads this means tool calls must be auditable end-to-end. Which tool was called, with which arguments, in what context, and producing which output — all traceable per decision. Without that record, the team cannot defend an automated decision to an examiner or reconstruct an incident.
In practice
Prism Agent Trajectories captures every tool call in a trace alongside the LLM call that triggered it, the reasoning that led to it, and any memory access. Per-step input, output, latency, and success status are recorded so reconstructing why an agent invoked a particular tool takes minutes, not days.
Related
More glossary terms
Start tracing in 5 minutes
One SDK. Five minutes. Full audit trails, PII redaction, and guardrail enforcement, from day one.