Model Context Protocol (MCP): The New Standard Every AI Developer Should Know

In case you missed it, AI is here. 

Artificial intelligence is becoming more ubiquitous every day. It’s powering your daily search engine query results, detecting fraudulent activity on your credit cards, navigating complicated traffic situations while you’re in your driverless rideshare car, and it even created a viral video of bunnies jumping on a trampoline that fooled folks who actively work in this space (or so I’ve heard…). 

Despite how it may seem, AI can’t do everything. One thing, perhaps the most meaningful and important thing, that humans naturally gather and consider that AI cannot? Context. 

The ability for artificially intelligent systems to understand not only a request, but also the broader situation around that request, determines how useful and reliable the outcome will be. How can we provide context to these tools so they are useful and reliable? 

Enter Model Context Protocol (MCP), a new standard designed to address this challenge. 

What is Model Context Protocol (MCP)?

Model Context Protocol is a new standard that helps AI systems talk to each other and to the apps, data, and tools they rely on. Without the protocol and registered tools available, there are infinite things the Large Language Models (LLM) could do to get more context. Without MCP, you would have to create a hand-coded solution for an LLM to make calls to an API of some sort. 

With an MCP server, you can build an application that connects a LLM to external tools and data. This allows you to create agents that can complete tasks like managing data and answering complex questions. 

Below are some key points about MCP: 

• MCP (Model Context Protocol) is a connection protocol designed to let applications, tools, and LLMs communicate in a structured way.

• It defines how clients (like an IDE, CLI, or UI) and servers (like a data source or API) exchange structured messages.

• Its purpose is standardization. Instead of every integration being custom, MCP provides a consistent format for capabilities like prompts, resources, tools, and events.

Credit Greg Isenburg on YouTube

MCP vs. API: What’s the difference?

You may hear MCP servers compared to APIs for the sake of a simple explanation of this newer concept. From a high level, this is a good comparison for general understanding. In simple terms: MCP helps artificially intelligent systems understand not just what is being asked, but also the bigger picture around the request. 

Let’s quickly re-establish what each of these is: 

• MCP: Think of it like a universal translator that makes different tools and apps speak the same language. A Rosetta Stone, if you will.

• API (REST, GraphQL): Think of it like a menu that shows you what data or actions a specific service makes available. Like a server in a restaurant — where you pick something off a menu and ask them to go get it for you. 

Now, how do they differ?

• REST APIs and GraphQL endpoints are about exposing business data or functionality over HTTP.

• MCP is not limited to HTTP — it can use different transports (e.g., stdin/stdout, sockets, or websockets).

• REST/GraphQL describe the what (business resources and operations), while MCP describes the how (a shared protocol for connecting clients and providers, including LLMs).

MCP is how the conversation happens. APIs are what you can talk about.

Why should your business care?

AI is powerful, but today, most tools are siloed. For example, an AI writing assistant might not know what is in your CRM. A customer service AI agent might not be able to see your store’s live inventory. That lack of connected context limits the impact AI can have on your business and for your customers.

MCP represents a shift in how AI will evolve: from isolated assistants to interconnected, context-aware systems that feel more natural and useful. In commerce, this could mean smarter product recommendations, more accurately responsive customer service, and new levels of personalization that directly drive results.

Organizations should view MCP as more than just a technical standard. It is the foundation for how AI will begin working across platforms, tools, and data sources. For developers, it is an opportunity to be at the forefront of what comes next in AI-powered commerce.

Think about it like this: if you ask your boss which item you should purchase from your cousin’s wedding registry, your boss totally lacks the necessary context to intelligently answer this question. However, if you granted your boss access to the registry and a data set about your cousin’s taste, they would have the context required to answer the question. 

MCP helps solve this context by:

• Making systems interoperable: AI tools can connect more easily and speak the same language.

• Improving accuracy: With shared context, AI delivers more relevant and reliable results.

• Creating seamless experiences: Both customers and employees benefit from AI that feels connected rather than fragmented.

For organizations, this means AI can move from being a set of isolated helpers to becoming an integrated part of operations. For developers, this means you can go from “What can we build?” to “What can’t we build?”

What can you build with an MCP? 

Right now, for many companies, MCPs are table stakes, and the ipso-facto example of implementation is creating a chatbot that interacts with your documentation, your store catalog, your flight-booking agent — and so on. Thinking about this, I found myself wondering: “Is a chatbot really that interesting?” 

The more people I spoke to - the more I realized: it isn’t the chatbot itself that is interesting. Rather, it’s how you leverage chatbot functionality to create interesting experiences and solutions. Your solution might look like a chatbot, but what makes your implementation special? What makes it unique? Therein innovation can prosper. 

In addition to that, in my research of what you could build with the use of an MCP server, I found a lot of cool examples. If you’re a developer, what would you add to this list? 

What can you build or do with an MCP server? 

• Customer service chat bot 

• Design assistant integrated with Figma or Canva APIs

• Content generator that creates marketing and blog content 

• An access management agent through natural language prompts 

• Analytics dashboards that respond accurately to natural language queries 

• Report generator 

• CRM/ERP Integrations that pull in sales, customer and inventory data 

• SO MUCH MORE!! 

What is Commerce doing with MCP? 

Our BigCommerce Storefront MCP is in development, and we are excited to welcome developers into our first (of many!) beta group at the end of September. 

For developers, MCP presents a unique opportunity. Think about the developers who sat at the bleeding edge of technology when the web first came out — that’s you, today, with MCP. 

This month we are launching the inaugural cohort of our AI Labs Developer Beta, an exclusive program that provides early, direct access to the BigCommerce Storefront Model Context Protocol (MCP). The group of developers we accept will be looked at as pioneers and consultants for our MCP implementation. 

The beta gives developers direct access to the BigCommerce Storefront Model Context Protocol, enabling the creation of new AI-powered commerce experiences. This is your chance to be among the first to shape the future of AI at Commerce.

This access allows developers to:

• Connect storefront data directly with AI tools

• Build AI-powered shopping experiences that feel seamless and personalized

• Experiment with new use cases for bringing AI into storefronts in ways not thought possible before

Ready to explore what MCP can unlock for your storefront? Apply to join AI Labs Developer Beta.