A travel-booking assistant built as a Foundry agent needs to tell customers today's lowest fare for a specific flight route. Fares change by the minute and the airline exposes a live pricing API. Embedding a snapshot of prices into the system prompt would go stale within hours. What is the best way to keep the assistant's fare quotes current?

The correct answer is registering the airline's pricing API as a tool the agent can call at conversation time, then having the agent invoke it and read the live result. That lets the assistant fetch the actual current fare at the moment a customer asks, so the answer reflects live pricing rather than a snapshot that grows stale within minutes. The retraining-the-model-nightly approach is incorrect because retraining a model nightly still leaves the assistant working from the previous day's numbers for the entire following day, which does not solve a problem that changes minute to minute. The morning manual system-prompt update is incorrect because a morning manual update still goes stale as soon as prices shift later that day, and it depends on someone remembering to do it. The larger-context-window approach is incorrect because a larger context window only lets the model hold more text at once; it does not make old fare data current, it just means more stale numbers fit in the prompt. Tool calling is the mechanism built for exactly this situation: instead of baking facts into the model or the prompt, the agent reaches out to a live system during the conversation and reads back a fresh result, which is the only approach here that actually guarantees the quoted fare is current.