Inside Anthropic’s Model Context Protocol (MCP): The New AI Data Standard

In November of 2024, Anthropic announced that it was making its Model Context Protocol (MCP) open source. The announcement calls this protocol, “A new standard for connecting AI assistants to the systems where data lives, including content repositories, business tools and development environments. Its aim is to help frontier models produce better, more relevant responses.”

The announcement has been met with widespread anticipation and support from the nerdier communities among us. But what is this standard? Where did it come from? Why is it needed? Who uses it? And why should you care?

Here is a quick explainer of Anthropic’s Model Context Protocol.

What Is Model Context Protocol (MCP)?

Model Context Protocol is a standard that defines how the details of your conversations with an AI — your language, time zone, past chats, etc. — are shared between AI models and the systems that use those large language models (LLMs). It provides a container — or structured system — to tell the AI model who it is talking to, what it is being asked to do, what you already know and what rules to follow — before it returns a response to your prompt.

Standards are often a sign that a technology is maturing. Bluetooth standards govern the way that technology is implemented. Wi-Fi standards define the way the hardware for that technology is built. HTTP standardizes communication via the internet and SMTP enables email transmissions. These technologies expanded and became ubiquitous once rules were established to help developers build tools that could talk to each other.

The MCP standard is meant to normalize communication between LLMs and whatever system is interacting with them. It acts like a translator between the raw data in the LLM and the input point where you enter your prompt, helping the AI to better understand the context around your questions. By giving them a common language for interpreting data, it will help chatbots, virtual assistants and other AI tools be less weird and provide better answers to questions.

It is, according to the news release, “an open standard that enables developers to build secure, two-way connections between their data sources and AI-powered tools.”

Where Did MCP Come From?

MCP was proposed and developed by Anthropic, the company that built the popular Claude LLMs. The company’s tools include HAIKU, a fast and light model; Sonnet, a hard-working LLM that is a useful blend of speed and high-throughput; and Opus, a high-performance model that can tackle complex analysis, coding and more challenging tasks.

You can ask Sonnet to build MCP server implementations for you or your organization if you aren’t a developer capable or interested in building your own. Anthropic is also sharing pre-built MCP servers for popular enterprise systems like Google Drive, Slack, GitHub, Git, Postgres and Puppeteer.

The company introduced the standard in early 2024 with the goal of helping all LLMs produce better and more relevant responses to queries. They made it open source with the goal of allowing the AI industry to expand.

“As AI assistants gain mainstream adoption, the industry has invested heavily in model capabilities, achieving rapid advances in reasoning and quality,” reads the announcement.

“Yet even the most sophisticated models are constrained by their isolation from data — trapped behind information silos and legacy systems. Every new data source requires its own custom implementation, making truly connected systems difficult to scale.”

It is often protocols, such as those developed for Wi-Fi, Bluetooth and the internet, that allow a technology — regardless of its developer — to gain traction rather than languishing in silos until going the way of the Betamax.

Why Is Model Context Protocl Needed?

Since ChatGPT launched to the public in 2022, people have flocked to AI for help with everything from simple questions to advanced research projects. They are increasingly becoming an important part of business, research and life.

But LLMs aren’t aware. They don’t have a memory. They rely heavily on the details of the prompt to understand what might seem obvious to a human: Who they are talking to, the goal of the question or work, the tone of the conversation and much more. Part of prompt engineering is hard coding this sort of context into each prompt or application.

The goal of MCP is to give the LLM some of that consistent, reusable context so that writing prompts is easier and less prone to weird, off-the-wall responses from the model.

MCP is like a shared vocabulary that allows the agents to understand context.

Remember when every camera, phone, headphone and smart home appliance came with its own charger and could not connect to your computer? That’s where AI is now — in a more abstract way. MCP is like the USB-C plug that will make everything work together, connect more easily and operate as a connected network.

Who Uses MCP?

If you are a developer, you are probably happy about this protocol. “Thanks to the Model Context Protocol (MCP),” said Derek Ashmore, AI enablement principal at Asperitas, in an opinion piece on CIO. “DevOps teams now enjoy a litany of new ways to take advantage of AI. MCP makes it possible to integrate AI into a wide variety of common DevOps workflows that extend beyond familiar use cases like code generation.”

It is the standard that will allow developers who want to integrate AI into the tools they are building to do that.

“Whether you’re building an AI-powered IDE, enhancing a chat interface or creating custom AI workflows, MCP provides a standardized way to connect LLMs with the context they need,” according to the MCP document specifications.

Why Should You Care About Model Context Protocol?

If you are a developer or work in an enterprise that uses AI, this protocol will help to unlock smarter AI agents that are more aware of the context around the work they do, need less coding and are faster to build and implement.

Having a standard for how the LLM accesses your data and other sources of information will make chatbots and AI agents smarter, less prone to unusual tangents and more able to understand what is an appropriate answer or action.

Learning Opportunities

Webinar