AI Context Hub
| The AI Context Hub is available as a beta feature. As the toolset continues to develop, you may notice updates and improvements over time. |
The AI Context Hub is a Neptune DXP - Open Edition toolset that catalogs a connected SAP system and lets AI agents work with it in natural language through governed SAP data sources. A request is translated into the SAP operations needed to carry it out, so no custom integration has to be built or maintained for each one. A data source can read from the connected system and, where it is configured to, create, update, and delete data. It runs against an existing SAP system, including on-premise ECC, and needs no SAP cloud contract or SAP BTP.
| Read access is the common case. Create, update, and delete operations are supported, but they change data in the SAP system, so grant and use them with caution. |
You configure an SAP data source in the SAP Data Sources tool. It decides which parts of the SAP system an agent may reach and how much of the data the AI model is allowed to see. An agent in Naia Agent Studio then uses the data source to turn a request into the right sequence of SAP calls, run them, and return a readable answer.
The AI Context Hub connects to SAP through Neptune DXP - SAP Edition and an imported bridge transport, and needs the API Factory access configured on the SAP side. Confirm these before you start. See System requirements.
Catalog-based request processing
The AI Context Hub answers from a catalog of the connected SAP system rather
than from the system directly. During ingestion, the orchestration engine records
the system’s services, operations, and fields, including custom objects, in a
searchable catalog. For each request, it searches the catalog for the matching
operations, plans the calls and their parameters, runs them against the live
SAP system as the configured user, and returns the result. The AI model configured
on the data source performs the planning. For example, the request Which vendors
have the largest open purchase orders? resolves to the operations that look up
vendors, read purchase-order values, and rank them. See How the AI Context Hub works.
Result formatting
The catalog records how the SAP system represents its data, and the orchestration
orchestration engine applies this to every result: codes are resolved to names, and amounts
are formatted with their currency. A result appears as ACME Corp and $150.00
rather than as a vendor number and an unformatted figure.
Supported source types
A single data source can reach OData services, BAPIs and Business Objects, search helps, standalone function modules, Neptune DXP - SAP Edition APIs, and SAP Integration Hub APIs. Any non-SAP API described by an OpenAPI 3.x specification is supported as well. See Sources you can connect.
Access control
The data source, not the AI model, determines what an agent can reach. It is granted access to specific services and operations, and a separate setting determines whether the AI model may read actual values or works from metadata alone. See Data security and governance.
Applicability
The AI Context Hub discovers services during ingestion and plans each request at run time, which suits open-ended requests that are not known in advance. For a fixed, predefined operation, the Hub MCP Designer exposes selected SAP logic to an agent as a tool.