Kensho Solves Reliable Financial Data Retrieval with LangGraph-Based Multi-Agent Framework

Financial professionals spend a lot of time finding and verifying the information they need for analysis due to fragmented data systems. To solve this, Kensho developed the Grounding framework, which provides trustworthy insights based on S&P Global's verified data.

Grounding is a multi-agent framework that uses LangGraph as its engine. A central router analyzes users' natural language queries and intelligently routes them to specialized Data Retrieval Agents (DRAs) for fields such as equity research, fixed income, and macroeconomics.

This structure separates the data routing layer from the data retrieval layer, allowing each team to manage its agents independently. The router combines the distributed responses received from multiple DRAs to provide users with accurate, context-appropriate, unified information.

In addition, a custom DRA Protocol was introduced for consistent data exchange. This protocol defines a common data format that includes both structured and unstructured data, enabling various financial AI products to be deployed quickly on the same robust data foundation.