Model Context Protocol

How MCP became enterprise infrastructure, how to build governed MCP servers, and why standardizing on the protocol still needs a semantic layer behind it.

The Model Context Protocol (MCP) is the open standard, introduced by Anthropic in November 2024 and now governed by the Linux Foundation's Agentic AI Foundation, for connecting AI applications to tools, data, and systems. It collapses the M-by-N integration problem into M-plus-N: write your system as one MCP server, and any MCP client can reach it.

This collection covers MCP from the enterprise architect's point of view. Posts here examine the adoption evidence and governance model that make MCP a real infrastructure decision, the exact wire shapes and code for building a governed MCP semantic layer server, and the semantic gap the protocol leaves open. MCP standardizes access. It does not standardize meaning. These pieces argue that the winning enterprise pattern pairs MCP as the transport with a governed semantic execution layer as the logic.

← Back to all posts

MCP as a unified access layer connecting public data sources and enterprise data systems to AI agents, with standard access, governance, identity, security, and observability.

Model Context Protocol 14 Jul 2026

MCP for Public and Enterprise Data: The New Standard for AI-Ready Access

The public MCP servers, the OpenAPI-to-MCP pattern, and the two-axis map of what AI-ready access really means across public and enterprise data.

AI agents connect through the MCP protocol layer to a governed semantic layer that provides metrics consistency, business meaning, policy enforcement, and deterministic SQL, then to enterprise data systems.

Model Context Protocol 14 Jul 2026

MCP Is Not Enough: Why Enterprise AI Agents Need a Governed Semantic Layer

MCP won the transport war. The dbt 2026 benchmark, the Gartner projection, and why meaning, not connectivity, is where production agents fail.

An AI agent routed through a governance control hub of identity, policies, data classification, audit, and enforcement before reaching MCP-connected databases, APIs, files, and data services.

Model Context Protocol 14 Jul 2026

Who Controls What an AI Agent Can Ask? Governance in an MCP World

The confused-deputy problem, the six layers of MCP governance, and why row and column authorization belong at compile time, not the endpoint.

AI agents and business users connect through an MCP layer to a governed semantic layer, then to enterprise data sources, producing trusted BI outputs.

Model Context Protocol 14 Jul 2026

MCP for Business Intelligence: How AI Agents Should Query Enterprise Data

Text-to-SQL demos work; production breaks with wrong joins and metric drift. The dbt 2026 benchmark, the Spider 2.0 cliff, and the governed semantic layer that fixes it.

A five-stage progression showing MCP moving from developer experiments to a standardized, governed connection layer for enterprise AI.

Model Context Protocol 14 Jul 2026

Enterprise MCP Adoption: Why the Model Context Protocol Became Infrastructure

The Linux Foundation donation, the version timeline, cross-vendor support, and why standardizing on MCP still needs a governed semantic layer.

APIs connect applications to systems, MCP connects agents to tools and data through an MCP server, and A2A connects agents to other agents, composing into one enterprise AI stack.

Model Context Protocol 14 Jul 2026

APIs vs MCP vs A2A: What Enterprise AI Teams Need to Know

A cited comparison of APIs, MCP, and A2A, a three-question decision framework, and why none of them solve the meaning problem.

Agent clients converge on one MCP wire into a governed semantic layer that resolves intent, proves join paths, and governs, then compiles dialect-perfect SQL to warehouses.

Model Context Protocol Updated 11 Jul 2026

How to Build an MCP Semantic Layer Server (Architecture, Code, and the No-Rip-and-Replace Case)

The protocol in two minutes, the exact message shapes a semantic-layer server speaks, a buildable FastMCP scaffold, and the cost argument for MCP over a custom API.