Definition
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to outside tools and data. An MCP server describes the tools it offers, and an MCP-compatible AI decides when to call them while it works on a request.
In practice
MCP was introduced in late 2024 and has since been adopted by many AI clients. It separates the assistant from the systems it works with: the assistant handles the conversation, and each MCP server handles one or more outside systems through a common interface.
For marketing, that means an AI can pull a Google Ads report, read a Tag Manager container or check GA4 events directly, instead of describing where to click. The protocol itself does not decide what the AI is allowed to change; that depends on the server and its controls.
When evaluating an MCP setup for live accounts, the questions that matter are which tools are exposed, how credentials are handled, and whether write actions are held for human approval.
Learn more: What is MCP for marketing.
Related terms
- MCP serverAn MCP server is a service that exposes a set of tools, each with a name, description and inputs, that an AI assistant can call through the Model Context Protocol.
- Server-side taggingServer-side tagging moves tag processing from the visitor’s browser to a server container you control.
- Google tag (gtag.js)The Google tag, loaded as gtag.js, is a single tag that sends events to Google products such as Google Analytics 4 and Google Ads.
How Markitech helps
Markitech connects your AI to Google Ads, GA4, Google Tag Manager, Meta and 12 more platforms, so it can check how Model Context Protocol (MCP) is set up in your own accounts and tags and propose fixes. Every change is held until you approve it.