Perplexity adds Finance Search to Agent API to deliver live, cited financial data

Perplexity announced Finance Search as a new tool available through its Agent API that centralizes time‑sensitive financial data retrieval. The tool is built to supply live, provider‑backed figures and cited web sources in a single call instead of extracting answers from broad web text. Finance Search can return prices, company fundamentals, earnings call transcripts, analyst estimates and market context, with the goal of making finance answers more accurate, current and easier to verify for downstream use.

The product is surfaced through the Agent API: when a developer submits a request such as a valuation lookup, an earnings recap or a market monitor, the model can detect the need for financial data and route the request to the Finance Search tool behind the scenes. That single interface is meant to replace bespoke integrations for each licensed provider by normalizing outputs into a stable schema, so responses, charts and workflows receive consistent, structured fields regardless of which source supplied the figures.

Perplexity evaluated Finance Search on FinSearchComp T1, a time‑sensitive financial‑data retrieval benchmark measured after market close. Published results show Finance Search began with the highest accuracy for live financial data and remained the most consistently accurate configuration on that benchmark over time. Perplexity attributes this performance to retrieving structured, provider‑backed data rather than extracting answers from broader web text. The company also reported cost comparisons for the same FinSearchComp T1 queries. Because Finance Search returns live financial data directly, the model can operate on fewer, more relevant tokens instead of processing large volumes of web text; Perplexity says that configuration produced the lowest cost per correct answer among the cohort evaluated on the benchmark. Those comparisons are presented alongside the accuracy metrics in Perplexity’s published results.

Finance Search includes controls intended to increase developer trust and auditability: every result carries inline citations that show which source produced a figure and which specific item the model used. Developers can select which model to run, view token usage, and configure Finance Search for their application; the documentation lists the configurations that performed best in Perplexity’s benchmark as tested defaults. Perplexity positions Finance Search as the retrieval layer for live financial workflows, and notes that finance teams can use it with tools such as Computer for Professional Finance to build tearsheets, market monitors, earnings recaps and research memos that include current, cited data. Documentation is available for teams that want to begin integrating Finance Search into agent workflows.