adaptive retrieval

Flexible data access for complex, mult workflows

Submit a natural language query in any LLM or multi-agent system, and Adaptive Retrieval dynamically searches and routes across disparate S&P Global data sources to assemble accurate, cited results. Designed for complex workflows requiring data from a variety of sources.

● how it works

Adaptive Retrieval for your AI workflows

Adaptive Retrieval is an AI data retrieval solution that integrates with any AI or agentic application. Users or agents ask a natural language question, and a search agent directs queries to the right data retrieval agents to return the most relevant result. Answers return citations to S&P Global documents.

● Key use cases

Trusted S&P Global data, wherever your research takes you

01

Deep research

Retrieve analyst commentary and executive insights across earnings transcripts, company profiles, and S&P Global documents through open-ended natural language queries.

02

Multi-step analysis

Handle complex, multi-dataset questions automatically, from tracing deal activity across sectors to synthesizing competitive landscapes.

03

Autonomous agent workflows

Route queries dynamically across S&P Global data sources, enabling agents to retrieve, combine, and return cited results without manual orchestration.

04

Report generation

Populate business reports, summaries, and dashboards with trusted, cited S&P Global data.

05

Industry and company screening

Identify companies meeting specific financial or operational criteria by querying multiple datasets in a single request.



● Benefits

Grounded results for multi-step queries

Dynamic routing

Automatically searches and routes each query across S&P Global's datasets. No manual mapping of queries to specific data sources.

Verified and auditable

Every response includes citations back to the source dataset or document, so results hold up to scrutiny.

Handles complex queries

Decomposes a single natural-language question into sub-queries, each handled by a specialized retrieval agent, for multi-dataset requests that would otherwise take several manual steps.

Open-ended discovery

Built for exploratory research and open-ended questions where exact data sources aren’t known upfront, unlike fixed API calls.

Built-in orchestration

Removes the engineering burden of building custom orchestration logic to connect a query to the right dataset or endpoint.

Agent-ready

Integrates into any MCP-supported AI or agentic application, with native support for Claude and Gemini Enterprise.


● Key capabilities

Built for builders

Specialized data agents

A search agent deconstructs each query and routes sub-tasks to dedicated retrieval agents for financials, transcripts, ratings, and energy research.

Multi-technique retrieval

Combines vector search, tool calling, Text-to-SQL, and code generation, selecting the right technique for each part of a query.

Source-linked answers

Every result cites the underlying S&P Global dataset or document, so outputs are traceable and auditable.

MCP server

Connects to any MCP-compatible application, including Claude, through Kensho's remote or local MCP servers.



● Get started

Ready to see it in action?

Request a conversation with our team or find more details on S&P Global Marketplace.