MCP Model Context Protocol: Definition and Examples

The Model Context Protocol (MCP) is an open standard that allows AI models to connect to external data sources, tools, and services via a unified interface, eliminating the need for custom integrations for each source.

Full definition

The Model Context Protocol (MCP) is an open protocol developed by Anthropic that standardizes how artificial intelligence applications interact with external data sources and tools. Concretely, MCP acts as a "universal USB-C port" for AI: instead of creating a specific integration for each service (database, API, file system), MCP provides a unique, standardized interface that any tool can implement.

The MCP architecture is based on a client-server model. The AI application (the MCP client) connects to one or more MCP servers, each exposing specific capabilities: tools (executable functions), resources (queryable data), and prompts (predefined query templates). This separation allows the same model to simultaneously access a PostgreSQL database, a messaging service, a web browser, or any other system, without modifying its source code.

One of the major advantages of MCP is its growing ecosystem. Dozens of MCP servers are already available for popular services like GitHub, Slack, Google Drive, SQL databases, or web search tools. Developers can also create their own MCP servers in Python or TypeScript, making any internal system accessible to an AI agent.

For prompt engineering, MCP fundamentally changes the game. Instead of manually copying and pasting contextual data into a prompt, the model can directly query relevant sources in real time. This allows for richer, more accurate, and more grounded interactions, while reducing hallucinations due to lack of context.

Etymology

The term "Model Context Protocol" was coined by Anthropic when announcing the protocol in November 2024. "Model" refers to the language model, "Context" highlights the main goal — enriching the context available to the model — and "Protocol" indicates that it is a standardized communication standard, like HTTP for the web.

Concrete examples

A developer connects Claude to their internal database via an MCP server to analyze data in real time

Analyze last quarter's sales in our database and identify the main trends. Compare with the previous quarter.

An AI agent uses MCP to interact with GitHub: create issues, read code, and submit pull requests

Read the file src/auth.ts in the backend repo, identify potential security vulnerabilities, and create a detailed GitHub issue for each issue found.

An AI assistant enriches its responses by conducting web searches via an MCP search server

Search for the latest advances in quantum computing published this week and write a structured summary with sources.

Practical usage

In prompt engineering, MCP enables designing prompts that leverage real-time data rather than relying solely on the model's fixed knowledge. Concretely, instead of manually including data in your prompt, you can ask the model to directly query your sources via the available MCP tools. This opens the door to complex workflows where a single prompt can trigger a chain of actions: search, analyze, transform, and act on external data.

Related concepts

Function CallingTool UseAPI (Programming Interface)Retrieval Augmented Generation (RAG)AI AgentGrounding

FAQ

What is the difference between MCP and classic function calling?

Function calling is a native capability of language models to call predefined functions. MCP goes further by standardizing how these functions are discovered, described, and invoked. With function calling, each integration is ad hoc. With MCP, a server exposes its capabilities via a uniform protocol, allowing any compatible client to use them without specific configuration. MCP is essentially the standardization layer on top of function calling.

Do I need to know how to code to use MCP?

No, not necessarily. Many pre-built MCP servers are available and can be installed in a few commands. Tools like Claude Desktop or Claude Code natively support MCP: you just need to configure the desired servers in a JSON configuration file. However, if you want to create your own MCP server for an internal service, basic knowledge of Python or TypeScript is required.

Is MCP specific to Claude or does it work with other models?

Although MCP was created by Anthropic, it is an open source protocol that any model or application can adopt. Several third-party editors and tools have already integrated MCP support, including IDEs like Cursor and Windsurf, as well as AI agent frameworks. The goal is to become an interoperable standard for the entire AI ecosystem, much like what LSP (Language Server Protocol) did for code editors.

How to use this prompt

Copy the prompt with the button above.
Paste it into ChatGPT, Claude or your favorite AI assistant.
Replace the bracketed variables with your details, then refine the result.

About Prompt Guide

Prompt Guide is a free library of 2500+ ready-to-use prompts for ChatGPT, Claude and other AIs, with guides to learn prompting and tools to build and optimize your own prompts.

Prompt library Learn prompting Prompt builder Prompt optimizer

More definitions

Million Token Context: Definition and Examples

Capacity of a language model to process up to a million tokens in a single request, enabling analysis of very large documents, codebases

Model Card: Definition and Examples

A model card is a standardized document that accompanies an AI model to describe its performance, limitations, potential biases, and conditions of use

Model Distillation: Definition and Examples

Model distillation is a compression technique where a smaller model (the student) learns to replicate the behavior of a larger and more performant model (the teacher).

Model Registry: Definition and Examples

A Model Registry is a centralized system for storing, versioning, and managing machine learning models throughout their lifecycle, from training to production deployment.

Multimodal: Definition and Examples

A multimodal AI processes multiple data types: text, image, audio, video. Discover GPT-4o, Claude 3, and Gemini, their capabilities and limitations.

Multimodal RAG: Definition and Examples

Multimodal RAG is an extension of Retrieval-Augmented Generation that allows an AI model to search and leverage information from sources

Get new prompts every week

Join our newsletter.