Model Context Protocol Specification enterprise integration for AU architectures

Why it matters for NETEVO

MCP is the current dominant tool-binding protocol for agentic AI. Any organisation integrating an LLM with external tools or data sources after late 2025 has either adopted MCP or has made a deliberate decision to use something else; the specification is the single most consequential agentic-AI interoperability artefact published in that period.

MCP is the protocol-level interface on which the Bounded SaaS / Unbounded Agents distinction is realised. The bounded SaaS layer publishes a tool surface; the unbounded agent consumes that surface through MCP. The distinction names the conceptual split; the specification defines the interface where the split occurs — which makes the distinction operationalisable rather than abstract.

A licence and governance note: the project is transitioning from MIT to Apache-2.0, and the copyright holder is Model Context Protocol a Series of LF Projects, LLC. MCP is now governed as a Linux Foundation project, not an Anthropic-controlled specification — a material fact for any organisation assessing vendor concentration risk under prudential operational-risk standards. A release candidate dated 2026-07-28 is locked; final publication is scheduled for 28 July 2026.