At TechEd 2025, SAP announced extensive support for the Model Context Protocol (MCP). This gives AI agents access to SAP data from all business applications. This applies not only to SAP AI agents, but also to third-party agents. Developers can use their favorite AI tools to work with SAP data and perform actions. This is an essential step in keeping developers connected to SAP, especially in the era of agentic AI.
At various conferences, we hear the same story over and over again: AI agents are the user interface of the future. Users will log in less and less to the various business applications; they will soon be giving AI agents commands via more central applications. This means these AI agents must be able to communicate with all the different business applications.
The role of MCP
This is where MCP comes in. With MCP, AI agents can communicate with each other, execute commands, and exchange data. Regardless of the application the user wants to use, an underlying AI agent can communicate with an SAP AI agent via MCP to perform tasks. This could be as simple as submitting a leave request in SuccessFactors, creating a PO in Ariba, or requesting a comprehensive ERP report in SAP S/4HANA.
MCP is therefore the equivalent of APIs, but for AI agents. All tech companies have embraced MCP as a standard for AI agents to communicate with one another. However, SAP is going one step further with the MCP Gateway.
Read also: SAP presents 15 Joule agents for finance, HR, and supply chain
MCP Gateway enables access to the entire IT landscape
The MCP Gateway announced by SAP during TechEd uses SAP’s Integration Suite. All APIs and Flows built into this Suite and linked to third parties can be converted into an MCP tool. This allows Joule agents (SAP AI agents) to communicate with those APIs via MCP.
In principle, MCP is not designed to work with APIs; it is intended to operate at a deeper level. The efficiency and speed of MCP are not optimal when it is API-based, but as an interim solution, it is certainly not a bad idea.
SAP therefore states that Joule agents can also work with legacy SAP systems and other outdated third-party business applications. There are a huge number of business applications that have APIs but will not support MCP in the coming years.
MCP Gateway also offers governance and security
Another advantage of the MCP Gateway is that it can also provide governance and security when organizations run all their integrations and MCP tools through the SAP Integration Suite. The question, of course, is whether they want to do that and how practical it is. But it provides a central tool for governance and security.
Ultimately, for organizations with a complex portfolio, it is more convenient to integrate via a gateway than via a series of point-to-point integrations, where it is easy to lose track of the big picture.
SAP offers broad native MCP support
SAP itself is fully committed to MCP and is trying to support it as broadly as possible in its platform. For example, the SAP HANA Cloud will receive full MCP support from Q1 2026. This will give Joule agents direct access to SAP’s database engine. Agents built with Joule Studio and AI Foundation will also have full access to the various data models.
A Joule agent can therefore answer a complex question with a broad set of underlying data and context. It can identify relationships between customers and suppliers, as well as geographical dependencies via spatial data. Much of the context data is stored in vector embeddings, so that meaning is captured and can be retrieved via semantic search.
Thanks to MCP support, third-party agents can also communicate with the Joule agent to access that data. Governance and security are also monitored by the Joule agent, ensuring that individuals only have access to the data they are entitled to.
This is a very valuable addition for developers, as they no longer need to understand SAP’s data model. They can give a textual command, and Joule will create and execute the complex query in the background. This saves a lot of time in understanding the underlying technology.
SAP offers clear knowledge graphs for developers and agents
Although it is not necessary to delve deeply into the underlying technology, it is sometimes useful to know which data and context belong together. SAP therefore automatically generates knowledge graphs from metadata in HANA Cloud. This makes it very clear which data belong together and what the relationship is. As a result, an agent or developer can quickly see, for example, which order belongs to which customer, or which shipping label belongs to which order.
Local MCP servers for use in IDEs
Developers will also be pleased that SAP is now making local MCP servers available for its SAP Build frameworks. This allows modern IDEs such as Cursor, Claude Code, Cline, and Windsurf to directly use SAP frameworks. It works with model-driven tooling such as the Cloud Application Programming Model, SAP Fiori elements, SAPUI5, and SAP’s mobile development kit.
These modern IDEs or IDE extensions allow you to write code using AI. The developer writes a prompt in plain language, and the AI generates the necessary code, makes adjustments, or fixes bugs. According to SAP, the MCP servers have been developed with appropriate “enterprise-grade governance” and the “clean core” principle, which means that SAP software is expanded in the right way rather than making adjustments to the core, which would make future updates difficult.
ABAP code gets its own MCP server and LLM
In the first half of 2026, there will also be an MCP server for ABAP code. This is the programming language on which the SAP ecosystem runs. The goal of the ABAP MCP server is to enable the generation of ABAP code with AI. In addition to ABAP, SAP also supports other languages such as JavaScript, but ABAP remains more popular at the core.
SAP is also introducing new LLMs trained on ABAP code. These are expected to be available next year and should make it even easier for developers to develop software on the SAP platform.
SAP opens the door to agents and developers
All in all, we can conclude that SAP is opening the door for developers to work more easily with SAP applications and the platform. MCP ensures that every application and every AI agent will soon be able to integrate with SAP, extract data, and perform actions.
While SAP is very open to developers, it is also trying to convince business users to adopt a best-of-suite strategy, primarily using SAP applications. This is somewhat contradictory, although SAP itself sees it differently.
We can only conclude that opening the platform and providing support for MCP are essential steps that SAP must take. This actually applies to every business software provider. The future lies in AI, and the UI of the future will undoubtedly be an application driven by AI agents.
Tip: SAP rebuilds Ariba completely: what will change in February 2026