Beroe links DataHub to MCP to power AI procurement data

Beroe has integrated its DataHub procurement intelligence platform with the Model Context Protocol, in an early deployment of the emerging standard for domain-specific data access in artificial intelligence systems.

The move links DataHub's structured market and supplier information to AI models through MCP. It gives procurement teams a new route to feed trusted data into generative AI tools and autonomous agents.

DataHub contains cost, supply and commodity market information that Beroe has compiled over several decades. The platform acts as a central repository that organisations use in category strategy, supplier evaluation and market monitoring.

MCP is an open protocol that defines a consistent way for AI systems to connect with external data sources and tools. It sets out how large language models and AI agents can request, receive and work with information from different enterprise systems.

Beroe said DataHub's MCP integration focuses on structured intelligence rather than document retrieval. AI agents can request datasets and interact with analytical functions that sit on top of the data.

The company said this approach supports AI systems that provide recommendations or decisions instead of text summaries. It also said it expects users to embed DataHub access inside copilots, chat interfaces and automated workflows.

AI decision data

Supriyo Mukhopadhyay, Chief Technology Officer for Beroe, said the main challenge for AI in procurement lies in the quality and structure of the data that underpins decisions.

"As procurement moves toward AI-assisted decisions, one of the biggest challenges is giving agents reliable, decision-grade intelligence they can act on. General AI models are fantastic linguists, but they don't actually know the spot price of copper in Chile or the specific risks brewing in Taiwan," said Supriyo Mukhopadhyay, Chief Technology Officer for Beroe.

Mukhopadhyay said DataHub's connection through MCP gives AI systems a defined layer of structured information and context from Beroe's datasets.

"By connecting DataHub through the Model Context Protocol, we are adding an intelligence layer to AI systems, providing hard data and domain context in a secure, structured, and future-proof way. That's the difference between a tool that just summarizes an email and an agent that can make a business decision. MCP is a foundational step in how enterprise AI will evolve, and we want our customers to be ready for it," said Mukhopadhyay.

Domain-specific use

The integration represents one of the first known uses of MCP against a specialised procurement data platform. Many early MCP examples have focused on general data access and tool orchestration.

Beroe said procurement functions can now link DataHub intelligence into AI workflows that track macroeconomic developments, supplier exposure, input cost changes and market shifts. It said this will support higher levels of automation in category management, supplier monitoring and scenario planning.

The company described the MCP link as a single connector that works across different AI systems. It said this reduces the need for teams to create and maintain separate integrations for each model or vendor.

Beroe also said MCP offers richer context for AI systems than traditional REST APIs. The protocol allows AI agents to understand the structure of DataHub datasets and the analytical tools that operate on them.

Security and governance controls within DataHub carry through into MCP-based access. This keeps the same permission models and audit trails in place at AI touchpoints.

Architecture and governance

MCP's design separates the way AI models work from the details of each data source. It treats back-end systems as tools that the AI agent can call in a standard way. Beroe said this makes new DataHub features available to AI workloads without separate integration work.

Customers can expose DataHub through MCP to internal or external AI models, subject to existing security rules. This model allows central technology teams to manage policies and access for multiple AI services.

The company said MCP support forms part of its wider work on an "intelligence stack" for procurement. It described this stack as a combination of machine-readable data, curated insights and AI-aligned architecture.

Beroe expects organisations that are rolling out generative AI copilots and autonomous agents in procurement to use the integration as a data backbone. It said this will underpin forecasting, supply risk monitoring and cost modelling across categories.

The firm plans further work on DataHub's analytical functions and AI-facing interfaces. It said those updates will become available automatically to AI systems that already connect to DataHub through MCP.