Every enterprise is being told to “add AI.” The pressure is real. But SAP systems were built for structured, transactional data — and every AI agent was built to process natural language. The result, for most organisations, is a brittle custom script that wraps an RFC call, occasionally breaks, and requires a developer every time someone wants to ask a slightly different question.
There is a better way. It is called the Model Context Protocol, and it is worth understanding now — before your competitors do.
What MCP actually is
MCP is an open standard published by Anthropic in late 2024. The simplest way to think about it: USB for AI. Before USB, every device needed its own proprietary connector. After USB, one cable worked everywhere. MCP does the same thing for AI agents and data sources.
Instead of writing a custom integration every time you want an AI agent to call a specific system, you build one MCP server for that system. Any MCP-compatible AI agent — Claude, GPT-4, Cursor, or a custom agent you build yourself — can then use it immediately, with no further integration work.
What this means for SAP
Your SAP system contains decades of business logic and data. Purchase orders, vendor master records, GL accounts, production orders, HR data, cost centre hierarchies. Without MCP, getting an AI agent to interact with any of this means writing custom code for each use case, managing SAP authentication, handling errors, and maintaining everything indefinitely.
With an SAP MCP server, you expose your SAP functions — OData service endpoints, RFC modules, BAPI calls — as tools that any AI agent can discover and call. The AI figures out which tool to use and when, based on what the user is asking. You build the connector once. Every use case after that is free.
A concrete example
A procurement manager opens their AI assistant and types:
“Which purchase orders from vendor 1234 are still open, and what's the total outstanding value?”
Without MCP: A developer writes a query, someone builds a report, it gets scheduled, and it answers that exact question — and only that question.
With an SAP MCP connector: The AI agent calls your MCP server, which executes the relevant BAPI or OData query against your live SAP system, and returns the answer in plain English. No new UI. No scheduled report. No ticket to IT.
The same agent, using the same connector, can handle completely different requests without any additional development:
- →Create a goods receipt for PO 4500012345
- →Show me all overdue vendor invoices from this quarter
- →What is the current stock level for material 100-110 in plant 0001?
- →Which cost centres overspent their budget last month?
Why not just wait for SAP Joule?
SAP Joule is real and it is coming. But it moves at SAP's pace — which means years of phased rollout, licensing tied to specific cloud modules, and limited ability to customise the AI behaviour to match your business processes. You are buying into SAP's AI roadmap, not your own.
MCP gives you model independence. You can use Claude today, switch to a newer model next year, and run custom agents trained on your own process documentation — all talking to the same SAP MCP connector you built once.
What your landscape needs
The good news: you do not need to replace anything. An SAP MCP connector sits alongside your existing system as a lightweight server layer.
- ✓Compatible with S/4HANA, ECC, and BTP
- ✓Works with any AI agent that supports MCP
- ✓Authentication via your existing SAP OAuth or basic auth setup
- ✓Data stays within your network unless you explicitly choose otherwise
- ✓No changes to your existing ABAP code or configuration
Is this production-ready?
MCP shipped as a stable standard in late 2024. Claude Desktop, Cursor, and a growing list of enterprise tools already have native MCP support built in. Early-adopter organisations are running production MCP integrations now. In 18 months this will be a baseline expectation.
The window to move first — and to do it thoughtfully, with the right architecture — is open right now. It will not be open for long.
Ready to connect your SAP system to AI?
We build SAP MCP connectors for enterprise teams. Tell us what you're trying to automate and we'll give you an honest assessment of what it would take.
Get in touch