What is an MCP server?

An MCP server is a program that makes some capability, like searching a database or creating a ticket, available to AI assistants through the MCP standard. The assistant reads the server's description of each tool and decides when to call it. MCP is the protocol and a server is one program that implements it, running either on your own computer or on a remote host.

What is the difference between MCP and an API?

An API is an interface developers write code against, one integration at a time. MCP usually wraps an existing API in a standard layer that any MCP-compatible assistant can discover and use without custom code, so in most cases it sits on top of the API.

What is the difference between MCP and RAG?

RAG is a technique for giving a model relevant documents before it answers, and MCP is a protocol for connecting a model to tools and data. An MCP server can supply the documents a RAG system uses, and it can also let the model take actions, which RAG doesn't cover.

Who uses the Model Context Protocol?

Anthropic created it. OpenAI announced support in March 2025, and in April 2025 Google DeepMind said it would support MCP in Gemini. In December 2025, MCP became a founding project of the Agentic AI Foundation, a Linux Foundation fund co-founded by Anthropic, Block, and OpenAI.

That depends on each server. The MCP specification asks host apps to get the user's consent before exposing data or calling a tool, but the protocol can't enforce that, so connect only to servers you trust, give them the narrowest permissions that work, and check what actions a server can take before you enable it.

Does ChatGPT support MCP?

Yes. ChatGPT can connect to remote MCP servers as connectors, and OpenAI's developer mode, which is in beta, adds full MCP client support, including tools that write data. Availability depends on your plan, so check OpenAI's help center for the current details.

How do you build an MCP server?

Use one of the official MCP SDKs, which exist for TypeScript and Python among other languages, to define what your server offers. Run it locally or host it remotely, then add it to an MCP client such as Claude or Cursor. The build guide at modelcontextprotocol.io walks through a first server step by step.

What Is MCP? Model Context Protocol Explained

Definition

Model Context Protocol (MCP)The Model Context Protocol (MCP) is an open standard for connecting AI applications to outside tools and data. An MCP server describes what it can do in a format any MCP-compatible assistant understands, so one integration can work across AI apps like Claude and Cursor without custom code for each.

How MCP works

The Model Context Protocol uses a client-server design with three roles.

Host: the AI app a person actually uses, such as Claude Desktop or the Cursor code editor.
Client: a connector inside the host that keeps a connection open to one MCP server.
Server: a program that exposes capabilities from an outside system like a CRM or an analytics platform.

Servers can offer three kinds of features. Tools are functions the model can call, like "search orders" or "create a calendar event." Resources are context and data for the user or the model to use, such as files and records. Prompts are templated messages and workflows the server provides for common tasks.

When a host connects, the server lists what it offers, with a plain-language description of each tool. The model reads those descriptions and decides when a tool would help with the request. Client and server exchange messages in JSON-RPC 2.0, a standard format for remote procedure calls.

It helps to think of MCP as a universal adapter. Without it, connecting 10 AI apps to 10 tools could take 100 custom integrations. With it, each tool builds one server and each app builds one client, and any client can talk to any server.

A short history of MCP

Anthropic introduced the Model Context Protocol as an open standard on November 25, 2024, with local MCP server support in the Claude Desktop apps. OpenAI announced support in March 2025, starting with its Agents SDK. In April 2025, Google DeepMind CEO Demis Hassabis said Google would support MCP in its Gemini models and SDK.

In December 2025, the Linux Foundation announced the Agentic AI Foundation, a directed fund co-founded by Anthropic, Block, and OpenAI, with MCP as one of its founding projects.

MCP vs API

	API	MCP
Built for	Developers writing code	AI assistants choosing tools
Integration effort	Custom code per app and per service	One server works with any MCP client
Discovery	Read the documentation	The server describes its own tools
Relationship	The underlying interface	Usually a layer that wraps an API

Most MCP servers call an existing API behind the scenes. What MCP adds is the standard description and connection layer, so an AI assistant can find that API and use it without a developer wiring it up first.

MCP vs WebMCP

The names are close and the jobs differ. MCP connects AI applications to servers, usually set up by the user or their company. WebMCP is a proposed web standard that lets an ordinary website expose actions like search or checkout to an AI agent working in the browser. A brand could use both, running an MCP server for customers' internal AI tools and WebMCP for agents that visit its site.

Why MCP matters for brands

MCP is one of the pieces that lets AI assistants act, on top of answering. Once an assistant can reach booking systems and online stores through a standard protocol, it can finish a task for someone rather than just recommend where to go. That's the foundation of agentic commerce, where agents research and buy on a shopper's behalf.

It also changes day-to-day work for marketing and SEO teams. Instead of exporting data from one dashboard into another, a team can connect its data sources to an assistant and ask questions in plain language.

Examples of MCP in use

A developer connects a code editor to a GitHub MCP server so the assistant can read issues and open pull requests.
A support team connects an assistant to its help desk, and the assistant looks up tickets and drafts replies.
A marketer connects Claude or ChatGPT to an analytics source and asks which pages lost traffic last week.

Searchable MCP is an example of that last case. It's a read-only Model Context Protocol server that lets AI assistants and automation tools query your Searchable data, including visibility, prompts, citations, sources, competitors, reports, and content opportunities. You can connect it to Cursor, Claude, ChatGPT, or n8n and ask something like "which domains does Perplexity cite most for our prompts?" without opening a dashboard. The launch post shows how teams use it in Claude Code and Cursor.

Frequently asked questions

MCP stands for Model Context Protocol. Anthropic introduced it as an open standard in November 2024, and it defines how AI applications connect to outside tools and data sources.

See how AI engines talk about your brand

Get a free report on where ChatGPT, Perplexity, and Google AI name and cite you.

Get your free visibility report

WebMCPA proposed web standard that lets a website expose actions, such as search or checkout, as tools an in-browser AI agent can call directly.Agentic commerceAgentic commerce is shopping in which an AI agent acts for a person: it finds and compares products and can complete the purchase, sometimes without the shopper opening the merchant's website. Open protocols such as the Agentic Commerce Protocol from OpenAI and Stripe and Google's Universal Commerce Protocol give agents and merchants a shared way to handle the order.Retrieval-augmented generation (RAG)Retrieval-augmented generation (RAG) is a technique where an AI model first searches a set of documents for passages that match a question, then writes its answer from those passages. Because it looks things up at answer time, the model can use up-to-date information and point to where it came from.

Large language model (LLM)An AI model trained on large amounts of text to predict and generate language. LLMs are the engines inside ChatGPT, Claude, Gemini, and most AI search tools.

GroundingGrounding in AI means tying a language model's answer to specific source material, such as live search results or a company's own documents, so the model isn't relying only on what it learned in training. A grounded answer can point to its sources, which is why AI search engines show citations.

Browse the full AI search glossary