An MCP server is a program that makes information and tools available to an AI application via the Model Context Protocol.<\/p>\n
The Model Context Protocol, or MCP for short, is an open standard for connecting AI applications to external systems. These may include files, databases, and executable functions.<\/p>\n
The architecture consists of three components: the AI application as the host, a client for communication, and the server, which provides information or functions. An MCP server can run locally or on a remote system.<\/p>\n
<\/p>\n
What opportunities does MCP offer businesses?<\/strong><\/h2>\nMCP connects AI applications with information sources and functions that lie outside the language model.<\/p>\n
One possible use case involves corporate chatbots that connect to multiple databases and can analyze data through a dialogue. The key factor here is the connection to a specific source.<\/p>\n
On the server side, MCP provides three building blocks: Resources provide context, such as content from files or databases. Tools execute functions, such as a database query or an API call. Prompts are reusable templates for interaction.<\/p>\n
When planning a use case, it is therefore important to make a clear distinction: What information does the AI need, and what action is it supposed to perform? Retrieving information from a system and making a change to that same system are different tasks.<\/p>\n
Here’s an example from our own experience: We developed an MCP server that powers our AI platform HeinzAI<\/a><\/strong> accesses the time tracking data in our SAP system. Employees can use it to view previously recorded times and enter new times \u2013 in other words, they can both retrieve information and modify data in the system.<\/p>\n![\"The](\"http:\/\/www.xeptum.com\/wp-content\/uploads\/2026\/10\/MCP-Server-Schmuckbild-1024x576.png\")
Through an MCP server, AI applications can retrieve information from enterprise systems and call defined functions.<\/p><\/div>\n
<\/p>\n
How does an available function become an AI tool?<\/strong><\/h2>\nA tool is assigned a description and defined inputs that the AI application can use to call it.<\/p>\n
An MCP server provides the available tools. The AI application can identify these tools and invoke them via the protocol. In an agent-based application, the language model can suggest an appropriate command based on the tool descriptions. The application controls the execution of the command and any necessary approvals.<\/p>\n
Among other things, the tool definition describes the function and the expected input parameters. The result is then returned to the application.<\/p>\n
The protocol governs the exchange. MCP itself does not specify how the application processes the received context or uses the language model.<\/p>\n
The importance of high-quality descriptions and inputs is also evident in our SAP time tracking system: For example, if someone asks about the hours logged the previous week, the tool description must clearly specify how the time period should be entered. Unclear descriptions can cause the language model to use incorrect inputs or fail to select an appropriate tool.<\/p>\n
Would you like to determine which functions of your business systems are relevant for an AI use case? We can help you analyze the process and develop a suitable integration architecture.<\/p>\n
Contact us now with no obligation<\/a><\/strong><\/p>\nWhat restrictions and controls are involved?<\/strong><\/h2>\nThe connection via MCP does not replace access controls or the verification of a tool call.<\/p>\n
MCP sets clear requirements here: Among other things, servers must validate input and implement appropriate access controls. Clients should obtain user confirmation for sensitive actions.<\/p>\n
However, the protocol does not specify which actions within a business process are technically permissible or appropriate. This must be determined for each specific use case.<\/p>\n
Therefore, for the intended use case, check which tools are provided, what types of input they accept, and at which points users are required to approve a request.<\/p>\n
In our own integration, this works as follows: Querying booked times is purely for informational purposes. A new time booking, on the other hand, modifies data in the SAP system and is therefore only executed after the employee has confirmed it.<\/p>\n
How can you prepare a suitable starting point?<\/strong><\/h2>\nA clearly described use case highlights the required data, functions, and decision points.<\/p>\n
First, describe the task, the systems involved, and the expected result. Then specify which steps retrieve information and which steps trigger changes. This provides the technical evaluation with a concrete business foundation.<\/p>\n
This is exactly where we at XEPTUM come in: We work with you to analyze AI use cases, develop suitable architectures, and implement the technical integration with SAP.<\/p>\n
Here\u2019s an example of how an MCP server could be used in your company: The sales department queries the AI to check delivery status and availability; the accounting department receives open items directly from the latest SAP data; and the purchasing department has purchase requisitions prepared, which are created in SAP only after approval.<\/p>\n
Our portfolio pages on SAP BTP & AI<\/a> and Business AI & Autonomous Enterprise<\/a> provide further insight into this interplay between processes, system integration, and AI. Our article on AI with SAP On-Premise<\/a> demonstrates how AI can be integrated into existing SAP systems.<\/p>\nXEPTUM is an SAP Silver Partner and has been helping companies further develop their SAP environments for more than 25 years. Learn more about our SAP partnership<\/a>.<\/p>\nWould you like to find out how AI can be securely and effectively integrated with your existing business systems? We can help you identify suitable use cases, evaluate technical requirements, and develop an appropriate integration architecture for your SAP and system landscape.<\/p>\n
Through an MCP server, AI applications can retrieve information from enterprise systems and call defined functions.<\/p><\/div>\n
<\/p>\n
How does an available function become an AI tool?<\/strong><\/h2>\nA tool is assigned a description and defined inputs that the AI application can use to call it.<\/p>\n
An MCP server provides the available tools. The AI application can identify these tools and invoke them via the protocol. In an agent-based application, the language model can suggest an appropriate command based on the tool descriptions. The application controls the execution of the command and any necessary approvals.<\/p>\n
Among other things, the tool definition describes the function and the expected input parameters. The result is then returned to the application.<\/p>\n
The protocol governs the exchange. MCP itself does not specify how the application processes the received context or uses the language model.<\/p>\n
The importance of high-quality descriptions and inputs is also evident in our SAP time tracking system: For example, if someone asks about the hours logged the previous week, the tool description must clearly specify how the time period should be entered. Unclear descriptions can cause the language model to use incorrect inputs or fail to select an appropriate tool.<\/p>\n
Would you like to determine which functions of your business systems are relevant for an AI use case? We can help you analyze the process and develop a suitable integration architecture.<\/p>\n
Contact us now with no obligation<\/a><\/strong><\/p>\nWhat restrictions and controls are involved?<\/strong><\/h2>\nThe connection via MCP does not replace access controls or the verification of a tool call.<\/p>\n
MCP sets clear requirements here: Among other things, servers must validate input and implement appropriate access controls. Clients should obtain user confirmation for sensitive actions.<\/p>\n
However, the protocol does not specify which actions within a business process are technically permissible or appropriate. This must be determined for each specific use case.<\/p>\n
Therefore, for the intended use case, check which tools are provided, what types of input they accept, and at which points users are required to approve a request.<\/p>\n
In our own integration, this works as follows: Querying booked times is purely for informational purposes. A new time booking, on the other hand, modifies data in the SAP system and is therefore only executed after the employee has confirmed it.<\/p>\n
How can you prepare a suitable starting point?<\/strong><\/h2>\nA clearly described use case highlights the required data, functions, and decision points.<\/p>\n
First, describe the task, the systems involved, and the expected result. Then specify which steps retrieve information and which steps trigger changes. This provides the technical evaluation with a concrete business foundation.<\/p>\n
This is exactly where we at XEPTUM come in: We work with you to analyze AI use cases, develop suitable architectures, and implement the technical integration with SAP.<\/p>\n
Here\u2019s an example of how an MCP server could be used in your company: The sales department queries the AI to check delivery status and availability; the accounting department receives open items directly from the latest SAP data; and the purchasing department has purchase requisitions prepared, which are created in SAP only after approval.<\/p>\n
Our portfolio pages on SAP BTP & AI<\/a> and Business AI & Autonomous Enterprise<\/a> provide further insight into this interplay between processes, system integration, and AI. Our article on AI with SAP On-Premise<\/a> demonstrates how AI can be integrated into existing SAP systems.<\/p>\nXEPTUM is an SAP Silver Partner and has been helping companies further develop their SAP environments for more than 25 years. Learn more about our SAP partnership<\/a>.<\/p>\nWould you like to find out how AI can be securely and effectively integrated with your existing business systems? We can help you identify suitable use cases, evaluate technical requirements, and develop an appropriate integration architecture for your SAP and system landscape.<\/p>\n
What restrictions and controls are involved?<\/strong><\/h2>\nThe connection via MCP does not replace access controls or the verification of a tool call.<\/p>\n
MCP sets clear requirements here: Among other things, servers must validate input and implement appropriate access controls. Clients should obtain user confirmation for sensitive actions.<\/p>\n
However, the protocol does not specify which actions within a business process are technically permissible or appropriate. This must be determined for each specific use case.<\/p>\n
Therefore, for the intended use case, check which tools are provided, what types of input they accept, and at which points users are required to approve a request.<\/p>\n
In our own integration, this works as follows: Querying booked times is purely for informational purposes. A new time booking, on the other hand, modifies data in the SAP system and is therefore only executed after the employee has confirmed it.<\/p>\n
How can you prepare a suitable starting point?<\/strong><\/h2>\nA clearly described use case highlights the required data, functions, and decision points.<\/p>\n
First, describe the task, the systems involved, and the expected result. Then specify which steps retrieve information and which steps trigger changes. This provides the technical evaluation with a concrete business foundation.<\/p>\n
This is exactly where we at XEPTUM come in: We work with you to analyze AI use cases, develop suitable architectures, and implement the technical integration with SAP.<\/p>\n
Here\u2019s an example of how an MCP server could be used in your company: The sales department queries the AI to check delivery status and availability; the accounting department receives open items directly from the latest SAP data; and the purchasing department has purchase requisitions prepared, which are created in SAP only after approval.<\/p>\n
Our portfolio pages on SAP BTP & AI<\/a> and Business AI & Autonomous Enterprise<\/a> provide further insight into this interplay between processes, system integration, and AI. Our article on AI with SAP On-Premise<\/a> demonstrates how AI can be integrated into existing SAP systems.<\/p>\nXEPTUM is an SAP Silver Partner and has been helping companies further develop their SAP environments for more than 25 years. Learn more about our SAP partnership<\/a>.<\/p>\nWould you like to find out how AI can be securely and effectively integrated with your existing business systems? We can help you identify suitable use cases, evaluate technical requirements, and develop an appropriate integration architecture for your SAP and system landscape.<\/p>\n
A clearly described use case highlights the required data, functions, and decision points.<\/p>\n
First, describe the task, the systems involved, and the expected result. Then specify which steps retrieve information and which steps trigger changes. This provides the technical evaluation with a concrete business foundation.<\/p>\n
This is exactly where we at XEPTUM come in: We work with you to analyze AI use cases, develop suitable architectures, and implement the technical integration with SAP.<\/p>\n
Here\u2019s an example of how an MCP server could be used in your company: The sales department queries the AI to check delivery status and availability; the accounting department receives open items directly from the latest SAP data; and the purchasing department has purchase requisitions prepared, which are created in SAP only after approval.<\/p>\n
Our portfolio pages on SAP BTP & AI<\/a> and Business AI & Autonomous Enterprise<\/a> provide further insight into this interplay between processes, system integration, and AI. Our article on AI with SAP On-Premise<\/a> demonstrates how AI can be integrated into existing SAP systems.<\/p>\n XEPTUM is an SAP Silver Partner and has been helping companies further develop their SAP environments for more than 25 years. Learn more about our SAP partnership<\/a>.<\/p>\n Would you like to find out how AI can be securely and effectively integrated with your existing business systems? We can help you identify suitable use cases, evaluate technical requirements, and develop an appropriate integration architecture for your SAP and system landscape.<\/p>\n