SAP and Anthropic announced the integration of Anthropic’s Claude into the SAP Business AI Platform and Joule, SAP’s AI assistant, using Model Context Protocol (MCP) standards to enable AI agents to coordinate across SAP S/4HANA, SAP SuccessFactors, SAP Ariba, and third-party systems. The partnership positions Claude as both a foundational model within SAP’s platform and as an extension of Joule’s orchestration layer, according to SAP CTO Philipp Herzig, who outlined the integration in a blog post this week.
The integration is designed to allow Claude-powered agents to execute tasks in finance, HR, procurement, and supply chain workflows without requiring custom integration code for each system. MCP provides a standardized protocol that allows AI agents to communicate with different business systems — SAP or otherwise — using a common language that any MCP-enabled system can understand.
Herzig said the integration reflects SAP’s approach that “AI needs to be carefully integrated into business processes – in a way that maintains and strengthens processes that already work.” Rather than restructuring organizations around AI, SAP and Anthropic are positioning Claude-powered agents to execute tasks within existing workflows, from quarterly financial closings to managing complex employee requests and rerouting supplier deliveries.
Claude Integration Across SAP Business AI Platform
MCP is a standardized protocol that allows AI agents to communicate with different business systems, SAP or otherwise, without needing custom integration code for each one.
Instead of building separate connections between Claude and every SAP application, database, or third-party tool it might need to access, MCP provides a common language that any MCP-enabled system can understand. Once implemented, this would allow Claude to request customer data from SAP S/4HANA, pull employee records from SAP SuccessFactors, check inventory in a supply chain system, or trigger an approval workflow in SAP Ariba – all using the same protocol.
Step by step, agents retrieve the data they need, make updates where authorized, initiate approvals through existing workflows, and complete tasks across multiple systems.
Analysis
What This Means for ERP Insiders
The integration relies on Model Context Protocol to connect Claude with SAP systems and third-party tools, which raises a practical question many enterprises overlook: who maintains those MCP integrations, and what happens when they break? If your procurement agent depends on live MCP connections between S/4HANA, Ariba, SuccessFactors, and external supplier portals, a broken integration doesn’t just slow the workflow — it can halt procurement decisions entirely. Before committing to agent-based automation, clarify with your vendor and SI partners who owns integration testing, monitoring, and support when agents span platforms.
How Claude and Joule Work Together in SAP Workflows
According to SAP Architecture Center documentation, Joule processes incoming user requests by analyzing three categories: the Scenario Catalog (metadata of all available scenarios, functions, and skills across SAP cloud applications), the Knowledge Catalog (SAP-owned and customer-owned knowledge), and user context and history including role-based permissions and chat history.
With Claude’s integration, SAP is adding Anthropic’s reasoning and agentic capabilities to this orchestration layer.
Claude-powered agents will be able to execute concrete actions for hundreds of thousands of SAP customers across finance, HR, procurement, and supply chain.
Analysis
What This Means for ERP Insiders
SAP positions this partnership as giving customers greater options through an open ecosystem, but deeper integration with Claude also means deeper dependency on Anthropic’s model roadmap, pricing, and availability. Tighter model integration usually means tighter vendor coupling, and that has cost, risk, and negotiating implications your procurement and architecture teams should evaluate now, not after deployment.
For example, a treasury manager could ask Joule to prepare a presentation for the CFO ahead of a meeting with a financial institution. Within minutes, Joule would deliver a completed presentation with current data, analysis, and notes on financial risks – tasks that previously required hours of manual effort.
SAP had announced extensive MCP support at TechEd 2025, giving AI agents access to SAP data from all business applications. SAP HANA Cloud added full MCP support in early 2026, giving Joule agents direct access to SAP’s database engine. All APIs and flows built into SAP Integration Suite and linked to third parties can be converted into an MCP tool, allowing Joule agents to communicate with those APIs via MCP.
Anthropic and SAP plan to collaborate strategically to develop custom agents and agent-supported workflows for key industries including public sector, healthcare, education, life sciences, and utilities, Herzig said.
Analysis
What This Means for ERP Insiders
When evaluating AI platforms that promise agentic capabilities, ask vendors explicitly where agent reasoning happens, what data the agent can access, and whether you can inspect and override actions before they execute in production. The distinction between “AI that recommends” and “AI that acts” is enormous, especially in regulated industries where audit trails and human accountability remain non-negotiable. Role-based permissions determine what agents can access, not when human oversight happens.
Editor’s note: A version of this story first appeared on SAPinsider.
