Enterprise MCP is the governed AI interface to Enterprise Knowledge. It packages business concepts, relationships, searchable context, available capabilities, permissions, and evidence requirements — so AI works through business language instead of raw systems.
It sits with Smart RAG and Viewpoints: assistants discover and request what the business allows, for whom, under Operational Trust — not by holding vendor credentials. Systems of record remain authoritative. Enterprise MCP is compiled after publish and certification — not a separate product you install beside Role Assistants.
Why it matters
AI tools need a stable, governed surface to discover context and request work. Enterprise MCP exposes certified enterprise capabilities and context through the Model Context Protocol — mediated by Operational Trust, not direct credentials to underlying systems.
How it works
Enterprise MCP combines:
- business resource types and metadata
- cross-system relationships
- search and retrieval context
- governed capabilities exposed for the active role
- permission and evidence requirements
AI clients connect to Enterprise MCP to understand what exists and which business actions may be requested. Execution still flows through capability checks, certification, and evidence capture. Enterprise MCP is one of two governed MCP layers — see Runtime Orchestration MCP for the universal Work-tool layer that sits alongside it.
What good looks like
Enterprise MCP is ready when business capabilities are published and certified, cross-system relationships are uploaded where Role Assistants need combined context, and roles are activated so users can reach the assistants that depend on it — one compiled, governed surface, not a checklist to memorize per system. For the exact publish → upload → certify → verify sequence and CLI checks, see the Enterprise MCP reference.
Enterprise MCP is the compiled surface after publish — not a separate product you install. If tools are missing, fix manifest, upload, and certification before expecting Role Assistants or external AI clients to call them.
Limits
Enterprise MCP tool catalogs depend on which systems are integrated, certified, and exposed in your environment. Not every capability or datasource appears in MCP until published through the integration lifecycle.
Example
An AI assistant uses Enterprise MCP to search customers, retrieve a governed customer overview, and request deal.reviewPipeline — each step scoped to the user's active role and certified datasources.
Business value
Enterprise MCP gives AI a business-language interface to the enterprise — without exposing raw APIs or bypassing governance.