MCP (Model Context Protocol) has been the talk of the AI world over the last couple of months, and it could end up becoming a fundamental part of AI infrastructure for years to come.
Kevin Scott, CTO of Microsoft, has articulated the potential impact of MCP during the recent Build conference, drawing a compelling parallel to the foundational HTTP protocol of the internet. He highlighted its simplicity and adaptability as key drivers of its growing adoption.
“The super exciting thing that we really want to talk about here at this Build is our real, serious commitment to MCP,” Scott said. :MCP has just taken off, like crazy, over the past handful of months, and there’s a good reason for it. It’s filling an incredibly important niche in this open web, open agentic web ecosystem,” he added.
According to Scott, the protocol’s strength lies in its minimalist design. “It’s a very simple protocol, akin to HTTP in the internet, where it allows you to do really sophisticated things just because the protocol itself doesn’t have much in the way of opinion about the payload that it carries, which also means that it’s a great foundation which you can layer things on top of.”
He then tied this to the evolution of the internet itself. “If you think about how the internet itself emerged, that composability of components and layering of protocols was really important to get to the richness. It gives you good starting places, and it gives you the ability to solve problems as you discover them as things are developing over time.”
Scott’s comparison of MCP to HTTP is significant. HTTP provides a standardized way for web browsers and servers to communicate, enabling the vast ecosystem of applications and services we use daily. If MCP achieves a similar status within the burgeoning world of AI agents, it could unlock a new level of interoperability and innovation.
The “agentic web” Scott refers to is a future where AI agents can seamlessly interact with each other and with various services, automating tasks and solving complex problems. For this vision to become a reality, a common language is needed – a protocol like MCP. The ability to “layer things on top” of MCP, as Scott mentions, is critical. This allows developers to build specialized functionalities and applications without reinventing the wheel each time. This layering concept is essential to scaling the internet, as the internet requires standardized protocols that are simple enough to be used in many different ways, depending on what the user needs.
Microsoft’s commitment to MCP aligns with recent industry trends emphasizing open-source and interoperability in AI development. For example, the rapid growth of open-source large language models (LLMs) and the increasing focus on AI agent frameworks reflect a desire for a more democratized and collaborative AI landscape. MCP could be a crucial component in facilitating this collaboration, allowing diverse AI agents developed by different organizations to communicate and work together effectively. This could foster innovation and accelerate the development of new AI-powered solutions.