Definition

Model Context Protocol (MCP)

Also known as: MCP, MCP server, Model Context Protocol server

The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems. An MCP server exposes tools, data and prompt templates that an AI assistant — such as Claude, ChatGPT, Cursor or VS Code — can discover and call, so the assistant can look information up and act on a user's behalf.

How it works

MCP follows a client–server model. The AI application is the host; for each MCP server it connects to, the host creates a client that keeps a dedicated connection to that server. Servers offer three core primitives: tools the assistant can call, resources it can read, and prompts that structure an interaction. Messages use JSON-RPC 2.0 over one of two transports — standard input/output for a local server running on the same computer, or Streamable HTTP for a remote one.

Why it matters

A vendor builds one MCP server and it works with every assistant that speaks the protocol, instead of a separate plug-in for each. For people who run fleets of equipment, that turns questions which used to mean navigating a management portal or calling support — which unit is offline, why, what to do next — into a conversation with the assistant they already use.

Common use cases

Asking an AI assistant for device status, telemetry and alarms across a fleet
Reading configuration settings without memorizing parameter paths
Pulling device logs and diagnostics into a troubleshooting conversation
Connecting an assistant to internal databases, ticketing systems or file stores

MCP is often described as a USB-C port for AI applications: one standard connector in place of a different cable for every device. The comparison holds for the people building integrations — and, increasingly, for the people using them, who no longer need a separate tool for every system they want to ask about.

MCP host, client and server: who does what?

Role	What it is	Example
Host	The AI application the user works in	Claude Desktop, Cursor, VS Code
Client	A connection the host opens to one server	One per connected server
Server	The program that exposes tools, resources and prompts	The DataRemote MCP

The assistant decides which of a server’s tools to call to answer a question, calls them through the client, and reads the results back into the conversation.

Local vs. remote MCP servers

A local server runs on the same computer as the assistant and communicates over standard input/output. Because it runs under the user’s own account, it keeps credentials and saved files on that machine. A remote server runs elsewhere and is reached over the protocol’s Streamable HTTP transport, with standard HTTP authentication. Neither is inherently safer — what matters is what the server exposes and what credentials it holds.

MCP vs. an API

A device management platform’s API is built for software integrations: provisioning systems, billing, ticketing. MCP is built for assistants. An MCP server translates between the two — it describes a set of tools in a way an AI assistant can discover, then calls the underlying API when the assistant uses one. That is why an MCP server can only ever do what the API and the user’s credentials allow.

How does MCP apply to POTS replacement fleets?

On October 7, 2026, DataRemote announced the DataRemote MCP, a local MCP server for its Ara device management platform. It lets an MCP-capable assistant answer questions about the POTS IN A BOX® appliances on a customer’s account — online status, LTE signal quality, FXS line registration, alarms, logs and configuration backups — and is read-only by default. See what it supports and how its controls work.

Common misconceptions

“Connecting an MCP server gives the AI unrestricted access to a system.”

A server exposes only the tools its author built, and it acts with the credentials it is given. Well-designed servers go further: the DataRemote MCP withholds every tool that changes a device or an account until an operator switches changes on in its settings.
“MCP is an AI model or an AI product.”

It is a protocol — a message format and a set of conventions. The intelligence lives in the assistant; the MCP server is the connector that gives it something to work with.
“MCP servers are always hosted by the vendor.”

MCP supports both. A local server runs on the user's own computer and talks to the assistant over standard input/output; a remote server is reached over HTTP. The DataRemote MCP is a local server.

Frequently asked questions

Who created the Model Context Protocol?

Anthropic introduced and open-sourced MCP on November 25, 2024. It is now maintained as an open project with a public specification and SDKs, and is supported by assistants and developer tools from several companies.

What is an MCP server?

An MCP server is the program that provides tools, resources and prompts to an AI assistant over the Model Context Protocol. It can run locally on the user's computer or remotely, and it typically wraps an existing system — a database, a file store, or a device management platform's API.

Is MCP the same as an API?

No. An API is a system's own interface; MCP standardizes how AI assistants discover and call tools. An MCP server often sits in front of an API — the DataRemote MCP, for example, works through DataRemote's API in Ara and never connects to devices directly.

Which AI assistants support MCP?

The protocol's documentation lists AI assistants such as Claude and ChatGPT and development tools such as Visual Studio Code and Cursor among its clients. Whether a given assistant can connect to a particular server also depends on the transport it supports — local or remote.