Guide

SAP AI Core MCP: community server, Generative AI Hub and limits

As of 2026-10-18

Edited by Cédric Mary, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

SAP AI Core MCP currently means one early-stage, MIT-licensed community server that exposes AI Core APIs as tools for managing a landscape in natural language. Our catalogue lists no MCP server for the Generative AI Hub, and no official SAP server for AI Core (18 October 2026). The two relate to MCP from opposite sides: AI Core runs models, MCP gives agents tools.

Looking for a SAP AI consultant? Cédric Mary, SAP AI & Analytics Architect & Team Lead, takes SAP Business AI, Joule and Business Data Cloud missions.

What the community server does

The server targets administration. It exposes AI Core APIs as tools and is driven by a single JSON configuration. The catalogue marks it early-stage and lists both transports. The practical use is operational: asking an assistant about scenarios, configurations, executions and deployments instead of calling the AI API by hand.

Our concept notes give the vocabulary the tools will follow: scenario, executable, configuration, then execution or deployment, with resource groups isolating runtime objects. Check that the server lets you pin a resource group, because it is the unit of isolation.

Generative AI Hub: not an MCP server

The Generative AI Hub is a capability of AI Core on the extended plan, with an orchestration scenario that offers a harmonised, OpenAI-style API in front of foundation models. It is where an agent gets its model, not where it gets its tools. We found no MCP server that wraps it, so a query for a Generative AI Hub MCP server has no verified answer beyond this relationship.

In an agent design the usual split is: the Hub or orchestration deployment serves the language model, and one or more MCP servers supply the tools and data. Both run through AI Core credentials on one side and the MCP server's own authentication on the other.

Other official servers that touch the platform

Two SAP-maintained servers sit nearby but are not AI Core servers. MCP Server for SAP BTP Administration exposes account-administration operations so an MCP-capable client can manage the account structure. SAP Automation Pilot MCP Server lets an agent run the Automation Pilot commands you select, which you expose as tools. Neither is documented as covering AI Core objects.

The practical reading for a platform team: AI Core administration through MCP is possible today only via a community server, while the surrounding BTP account structure and automation have official SAP-maintained servers. Mixing them in one assistant session is possible but doubles the number of credentials and scopes to review.

What to check before production

Credentials: an AI Core service key is a client ID and secret (or x.509 certificate) plus the XSUAA URL and the AI API URL, and rotation is the customer's job. A server holding that key can create deployments, which consume capacity units, so give it the least privilege you can and monitor spend. Scope: separate read tools from tools that create or delete deployments, executions and secrets. Audit: log tool calls against the resource group. Maturity: early-stage means interfaces may change; pin a version.

Also confirm the plan. The generative features need the extended plan, and the free plan ended on 22 May 2026 per our notes.

One more design question is whether you need MCP here at all. If the aim is only to call models, the orchestration deployment's API is the direct path and needs no MCP server. MCP pays off for repeated operational tasks, such as listing failed executions across resource groups, where a person would otherwise script the AI API.

Relation to Joule and agent platforms

Joule Studio can attach MCP servers to a Joule agent, so an agent built there can reach tools through MCP while its model access runs through SAP's own stack. An AI Core management server is an operations tool for platform teams, not something to give business users. For decisions on SAP AI staffing and rates, the Analytics Legends MCP offers read-only market data; it has no link to your AI Core tenant.

What we cannot assert

Which AI Core API groups the community server wraps, and its read-only options, are not stated in the catalogue.

Frequently asked

Is there an official SAP AI Core MCP server?

Not in our verified catalogue of 18 October 2026. The one listed is a community, early-stage server.

Is there a Generative AI Hub MCP server?

None found. The Hub serves models via AI Core's orchestration scenario; MCP servers provide tools to the agent that calls those models.

Can the assistant spend money through it?

Potentially: deployments consume capacity units. Restrict create tools and monitor consumption.

Is the server safe to expose remotely?

Both transports are listed. If remote, you own authentication, logging and rate limits; the service key it holds is powerful.

Engage the editor

This guide is written by Cédric Mary, SAP AI & Analytics Architect & Team Lead. 27 years of SAP — Datasphere, Business Data Cloud, SAC Planning, Joule. Freelance or permanent · Hybrid / remote · Anywhere in EMEA. Available from 19 October 2026.

Profile and contact

What this page is built on

SAP AI Core
SAP Generative AI Hub
MCP — Model Context Protocol in SAP (mechanics)
SAP AI Core MCP (community) — GitHub
MCP Server for SAP BTP Administration — SAP Help

External sources

Guides that answer with this page

These guides cite this page as one of the sources their answer rests on.