AI glossary

    What Is Model Context Protocol (MCP)?

    A plain-English definition of the Model Context Protocol: how MCP connects AI models to business tools, who governs it and how to use it safely.

    By Chad Westleigh · Updated

    Book an MCP scoping call See MCP integration services

    What is Model Context Protocol?

    The Model Context Protocol is a common plug standard between AI models and the systems they need to use. Before MCP, every connection between an AI assistant and a business system, such as a CRM, a file store or a database, had to be built separately for each assistant. Anthropic introduced MCP on November 25, 2024 as an open standard to replace those one-off integrations, and released it with pre-built servers for systems including Google Drive, Slack, GitHub and Postgres.

    MCP does not replace APIs. Most MCP servers wrap an existing API and describe it in a form a model can discover and use. The difference is covered in MCP vs API.

    How MCP works: hosts, clients and servers

    The MCP specification defines three roles that exchange JSON-RPC 2.0 messages:

    • Host: the AI application a person uses, such as Claude, an IDE or a custom agent.
    • Client: the connector inside the host that keeps a connection to one server.
    • Server: a service that exposes capabilities from a system, such as Shopify, Klaviyo or an internal database.

    Servers offer three kinds of features. Tools are functions the model can call, like "create a draft order." Resources are data the model or user can read, like a product catalog. Prompts are reusable templates for common tasks. Clients can offer features back, such as elicitation, which lets a server ask the user for missing information mid-task. As of October 2026, the current specification version is dated 2026-07-28.

    Who governs MCP?

    On December 9, 2025, Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation co-founded by Anthropic, Block and OpenAI, with support from Google, Microsoft, AWS, Cloudflare and Bloomberg. Anthropic said MCP's governance model would stay unchanged. At the time it reported more than 10,000 active public MCP servers and adoption in ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code. MCP is a cross-vendor standard, not a Claude-only feature.

    MCP vs API vs custom integration

    APICustom integrationMCP server
    Built forDevelopers writing codeOne app talking to one systemAI applications that discover and call tools
    Describes itself to a modelNo; needs docs and glue codeNoYes; tools carry names, descriptions and input schemas
    ReuseAny code that calls itUsually noneAny MCP-compatible host, such as Claude or ChatGPT
    Typical useSystem-to-system data exchangeA fixed sync or reportLetting an AI agent read and act in a business system
    Security modelAPI keys or OAuth scopesVariesThe system's own authentication, plus user consent in the host before tools run

    How Operator Intelligence uses MCP

    Operator Intelligence uses MCP to connect Claude and other models to client systems with narrow, auditable access. At ROAM Luggage, Claude makes requested site changes via MCP inside a Shopify development theme, and a team member reviews and publishes; critical updates went from days to minutes. Read the ROAM Luggage case.

    Our rules follow the specification's security principles: separate read tools from write tools, grant the minimum scopes, log every call, and put a person in front of anything that publishes, spends or contacts a customer. Tool descriptions from third-party servers are treated as untrusted until reviewed. Our MCP integration services page covers custom server builds, Claude Skills explains how instructions pair with MCP tools, and AI agent orchestration covers coordinating several tools and agents.

    MCP questions from business teams

    Is MCP only for Claude?

    No. Anthropic created MCP, but it is an open standard hosted by the Linux Foundation's Agentic AI Foundation, and products including ChatGPT, Gemini, Microsoft Copilot, Cursor and Visual Studio Code support it.

    Is MCP secure?

    MCP is as secure as its implementation. The specification says hosts must get user consent before invoking tools, but the protocol can't enforce good practice by itself. Use trusted servers, least-privilege credentials, separate read and write tools, logging and human approval for risky actions.

    When should a business build a custom MCP server?

    When a team wants AI assistants or agents to work in a system with no reliable MCP server, or when an existing server exposes more access than you want to grant. A custom server exposes only the tools and data a workflow needs.

    Does MCP replace Zapier or n8n?

    No. Automation platforms run scheduled, rules-based workflows; MCP gives a model a standard way to call tools during a task. They often work together: in the ROAM Luggage build, an n8n workflow feeds tasks to Claude, which uses MCP tools.

    Connect your systems to AI, safely

    Tell us which systems your team wants Claude or another model to work in. We'll scope the tools, permissions and review steps before anything is built. Or start with the AI readiness assessment.

    Book a scoping call

    Sources

    1. Anthropic, "Introducing the Model Context Protocol" (2024): Anthropic introduced the Model Context Protocol on 25 November 2024 as an open standard for secure, two-way connections between data sources and AI tools, with pre-built servers for Google Drive, Slack, GitHub, Git, Postgres and Puppeteer.
    2. Anthropic, "Donating the Model Context Protocol and establishing the Agentic AI Foundation" (2025): On 9 December 2025 Anthropic donated MCP to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation co-founded by Anthropic, Block and OpenAI, supported by Google, Microsoft, AWS, Cloudflare and Bloomberg; governance model unchanged; 10,000+ active public MCP servers; adopted by ChatGPT, Cursor, Gemini, Microsoft Copilot and VS Code.
    3. Model Context Protocol specification (version 2026-07-28) (2026): MCP uses JSON-RPC 2.0 between hosts, clients and servers; servers offer resources, prompts and tools; clients may offer elicitation; hosts must obtain explicit user consent before invoking any tool; tool annotations considered untrusted unless from a trusted server. Current version 2026-07-28.

    Get the Operator Intelligence briefing

    A short, practical note every two weeks on what is working with AI in real businesses. No spam; unsubscribe anytime.

    Related

    MCP vs API

    Why MCP sits on top of APIs, not in place of them.

    MCP integration services

    Custom MCP servers with scopes, logging and review gates.

    ROAM Luggage case study

    Claude managing a Shopify theme via MCP.

    Claude for business

    Projects, Skills, connectors and MCP for small teams.

    Klaviyo MCP and Claude integration

    An official MCP server, set up read-only first.

    AI glossary

    Sixteen AI terms defined for business leaders.