Analytics Legends logoAI & Analytics Legends Sign inSign up
Academy module

SAP's Developer MCP Servers for AI-Assisted Delivery: CAP, UI5, SAP Fiori and MDK in an Agentic IDE

SAP's Developer MCP Servers for AI-Assisted Delivery: CAP, UI5, SAP Fiori and MDK in an Agentic IDE — Analytics Legends section illustration for the SAP Analytics knowledge base (concepts, studies, Academy)

As of 2026-11-05

SAP publishes four open-source MCP servers that let an agentic IDE read project facts and call SAP tooling: CAP, UI5, SAP Fiori and MDK. This module reads their GitHub READMEs as the primary source. It lists the tools each server offers, the exact commands to register them in Claude Code and VS Code, the rules blocks that steer the model, and the behaviours that matter for safety: version pinning, first-use model downloads, saved SAP connections, TLS options, telemetry and logging. It separates these developer aids from the runtime MCP adapter of CAP and from the SAP API Policy question. You wire three servers into a project, record what worked on your processor and finish with a control list a security lead can accept.

What you will learn

  • Name the four SAP developer MCP servers and the tools each README lists
  • Register the CAP, UI5 and SAP Fiori servers at project scope in Claude Code with the README commands
  • Write a rules block that makes the model use the servers in a fixed order
  • List the documented network, credential, TLS and telemetry behaviours and the controls that follow
  • Separate developer aids from runtime agent access under the SAP API Policy

Concept overview

SAP's developer MCP servers are open-source programs that an agentic IDE starts on your machine so that its model can read project facts and call SAP tooling instead of guessing. Four of them are documented by SAP on GitHub as of 5 November 2026: the CAP server (cap-js/mcp-server), the UI5 server (UI5/mcp-server), the SAP Fiori server (@sap-ux/fiori-mcp-server in SAP/open-ux-tools) and the MDK server (SAP/mdk-mcp-server). Each README describes a server for AI-assisted development, and each one is launched by the MCP client over standard input and output, not over the network.

The decision this module trains has three parts. Which of the four servers does a given project need? How do you configure each in the agentic IDE the team uses (.mcp.json or the client's own file), with the exact commands the READMEs publish? And what must you restrict before the first prompt, because a server that starts with the developer's rights can read files, download models, call SAP systems and, in the MDK case, deploy? Concept cards C171 (Joule for Developers and agentic IDEs), C105 (adoption strategy for MCP) and C336 (ABAP AI) give the wider frame; this module stays on the four servers and their documented setup.

Prerequisites

  • Node.js and npm installed; the UI5 README requires v20.17.0 or v22.9.0 or higher
  • A CAP project and an MCP-capable client such as Claude Code, VS Code with Copilot or Cline
  • Runtime MCP and the SAP API Policy — see M327 and concept C105

Outcomes

  • Explain what each of the four servers reads or changes and which one a task needs
  • Reproduce the lab: three servers listed by claude mcp list and a rules block in CLAUDE.md
  • Pin a version and say why the Fiori README asks for it
  • Apply the five controls: pinning, telemetry, approval, sandbox systems, TLS
  • State what the READMEs do not document: support terms, compatibility matrix, permission model

Full module with the Consultant plan — €19.90/month. The full module adds: the decision framework · the end-to-end scenario walkthrough · the KPI scorecard · the anti-patterns · the code blocks · the knowledge check · the diagrams.

Open in the app →

Open missions

120 live roles and contracts in SAP AI & Joule right now — see the SAP AI & Joule missions · freelance SAP contracts.