The Parallel Decisions API is a web-grounded decisions engine that combines the speed of models such as Jev with live web recency in a single call. It is live at $0.001 per request with 5,000 free trial requests. Developers provide shared context and one or more typed questions (noul/predicate, choice or score) and receive probabilities, citations and sources. Batching shares retrieval. Example uses include fact-checking, refreshing banking records, classifying companies and leads, backtesting strategies and tracking competitor inventory.
Parallel’s Decisions API answers closed-ended questions using live web evidence. A request supplies shared context and named questions, each typed as yes/no (noul), choice, or score; up to 100 questions can share one retrieval. Responses include probabilities and citations, and the API can refuse when the available evidence is insufficient. The product is presented for workflows such as company and lead classification, vendor screening, and account monitoring. Developers can try it in Parallel’s playground or call it through the API; Parallel also documents access through its MCP server.
