Seamless access to S&P Global data across AI workflows

The S&P Global AI Data Portal enables seamless access to S&P Global data across AI workflows. It integrates with any MCP-enabled AI application, providing a single point of entry to S&P Global's trusted data. Two data access methods provide flexible access for your particular requirements. Deterministic Retrieval offers API-driven data access, with no AI-driven reasoning or generation. Customers retain full control over orchestration. Adaptive Retrieval builds on these capabilities. With a fleet of agents under the hood, this solution supports more complex or multi-agent workflows.

Deterministic Retrieval (also known as the Kensho LLM-ready API) delivers targeted, API-based access to S&P Global data. This approach works well for specific queries with clearly-defined parameters, and provides predictable, reproducible results. There is no AI-driven reasoning or generation as part of the solution, and customers have full control over orchestration. The solution is ideal for company-specific research, financial analysis, transcript intelligence, and multi-entity comparison use cases.

Adaptive Retrieval builds enables agent-driven data access for complex, open-ended research workflows. This method uses a fleet of agents under the hood to dynamically search across disparate data sources and handle multi-step queries and more complex analysis. The solution is ideal for deep research use cases, multi-step analysis, and autonomous agent workflows.

When you query a generative AI application, a model will generally reach into its memory or training data, search the web, or pull from a database it has access to. Then it generates an answer based on what it finds. That works well in many instances, but it’s also why models tend to hallucinate. It may not have all the context or domain expertise to understand the question. It may pull from inaccurate web data, or hallucinate an answer yet return it with utmost confidence.

Underneath every prompt, the S&P Global AI Data Portal chooses from a purpose-built set of tools and agents. Each one has been designed by subject-matter experts and engineers who understand the underlying datasets and how customers use them every day. These proprietary tools are the result of painstaking engineering effort and rigorous evaluation.

The tools and agents that underlie our solutions ensure that the model can accurately retrieve the right answer from S&P Global’s authoritative, trusted database, or send out a specialized agent to run a mini workflow and determine what’s needed to get you the right answer.

Every answer is grounded in S&P Global data with consistent, transparent results and source citations you can trace. Clients and compliance teams can verify exactly where data came from and how it was retrieved. Citations link directly into the S&P Global desktop. Click a citation and you’re taken to the exact paragraph in the 10-K, transcript, or research report; not just the document, the specific passage. For structured data, the underlying SQL query is surfaced alongside results. You see exactly what was queried, what filters were applied, and can reproduce the result independently.

Deterministic Retrieval is built for accuracy and efficiency, so your model analyzes only the data needed for the question without hallucinating or burning through its context window. This reduces the agentic overhead and helps you get what you need faster, often with fewer tokens.

The S&P Global AI Data Portal can be integrated into any MCP-supported AI application. Our MCP servers provide zero-configuration setup, automatic updates, enterprise-grade reliability, and performance across  AI platforms. Refer to our Adaptive Retrieval and Deterministic Retrieval (Kensho LLM-ready API) developer documentation for more details.

For faster out-of-the-box access, we’ve enabled native connectors with leading AI platforms. Adaptive Retrieval connectors are currently available within Claude and ChatGPT. Deterministic Retrieval supports native integrations with Claude, ChatGPT, Gemini Enterprise, Microsoft Copilot, Cohere, Amazon Quick Suite, Glean, and Databricks, with more on the way. We are working on connecting our solutions with additional partners across the AI ecosystem.

Kensho MCP Apps are currently in beta with selet clients. MCP Apps embed interactive S&P Global charts, tables, and visualizations inside AI chat interfaces like Claude and ChatGPT, turning static AI responses into live, explorable financial research. Clients can visualize, explore, and interact with S&P Global data inside AI applications. MCP Apps are served through iframes and sit outside the model context window, meaning they accrue a fraction of the token cost to develop embedded visualizations compared to LLM generation, with no risk of hallucinations. Learn more in our blog post.