MCP and Innovation Paradox: Why Open Standards will save AI from itself

Big model AI is not running the next wave of innovation. The original disruption is calm: standard.

Launched by Anthropic in November 2024, the model context protocol (MCP) standards how AI applications interact with their training data with the world. Most of the HTTP and REST standard web applications are connected to services, MCP standards how the AI ​​models connect to tools.

You have probably read a dozen articles in which it is said what the MCP is. But most miss boring – and powerful – part: MCP is a standard. The standards do not just manage technology. They make growth fly wheel. Adopt them quickly, and you are riding a wave. Ignore them, and you are behind. This article explains why the MCP is now important, what challenges are facing, and how it is already changing the environmental system.

MCP How does Move us from chaos to context

Meet Product Manager Lily at Cloud Infrastructure Company. She awakens projects like half a dozen tools such as Jira, Figma, Gut Hub, Slack, Gmail and Sangam. Like many people, she is also drowning in refreshments.

By 2024, Lily found out how good the LLMS (LLMS) had become good at combining information. She saw a chance: If she could open all her team’s tools in a model, she could draft updates, communication and answer questions on demand. But each model had its traditional way to connect with services. Each integration pulled it deep into the same vendor platform. When he needed to drag the copies from Gong, it meant that another basepock connection was to be created, which would be difficult to turn into a better LLM later.

Then Anthropic launches MCP: How an open protocol to make the standard flows into the LLM in context. MCP picked up baking quickly Open IFor, for, for,. AWSFor, for, for,. AzureFor, for, for,. Microsoft Co -Capital Studio And, soon, Google. Official SD’s available DearFor, for, for,. Type scriptFor, for, for,. JavaFor, for, for,. C#For, for, for,. RustFor, for, for,. Kotlin And Swift. For the Community SD Barley And others followed it. The adoption was fast.

Today, Lily runs everything through Claude, which is connected to its work apps through a local MCP server. Status reports produce itself. Leadership updates are a gesture away. As new models appear, they can change them without losing any of their integration. When she writes the code on the side, she uses the cursor and the same MCP server with an open model as she does in the cloud. Its IDE already understands the product she is building. The MCP made it easy.

A quality strength and implications

Lily’s story shows a simple truth: no one likes to use scattered tools. No user likes to be locked in shopkeepers. And no company wants to rewrite the integration whenever the model changes. You want the freedom to use the best tools. Provides MCP.

Now, the implications come with standards.

First of all, a strong public API is at risk of being obsolete. MCP tools rely on these APIs, and users will demand support for their AI applications. There are no excuses, with de facto standard emerging.

Second, the AI ​​application development cycle is going to be dramatic. Developers no longer have to write custom code to test simple AI applications. Instead, they can connect MCP servers with the readily available MCP clients, such as cloud desktop, cursor and wind surf.

Third, the switching cost is falling. Since the integration is duplicated with specific models, organizations can migrate to the open from the cloud to the open – or the models – or migrate to the infrastructure. Future LLM providers will benefit from the existing ecosystem around the MCP, which will focus on better price performance.

Navigating challenges with MCP

Each standard introduces new friction points or does not solve the current friction points. The MCP is also not exempt.

Trust is importantThousands of community -retained servers have been offered in dozens of MCP registrants. But if you do not control the server – or trust the party that does – you are at risk of leaking secrets to an unknown third party. If you are a sauce company, provide the official server. If you are a developer, find government servers.

The quality is variable: APIs are ready, and poorly maintained MCP server can easily be out of sync. LLMS rely on high quality metadata to find out which tools use. There is no authentic MCP registry yet, which strengthens the need for government servers of reliable parties. If you are a sauce company, keep your servers as soon as your APIS is ready. If you are a developer, find government servers.

Large MCP servers increase costs and low utility: Many tools in the same server increase costs through token consumption and many more choice models. If they have access to a lot of tools, LLMs easily confused. It is the worst in both worlds. Small, task -based servers will be important. Keep it in the construction and distribution of servers.

Permissions and identification challenges retain: These problems were present before the MCP, and they are still with the MCP. Imagine that Lily gave the capacity to send an email to the cloud, and give instructions well with intentions: “Send the status update to Chris quickly.” Instead of emailing his boss, Chris, LLM emailed every person named Chris in his contact list to get a message to Chris. Humans will need to stay in the loop for the highest steps of the decision.

Are looking forward to

MCP is not hype – this is a fundamental change in the infrastructure of AI applications.

And, just as it has before the standard of adapting before, the MCP is creating a reinforced fly wheel: every new server, every new integration, every new application mixes this pace.

New tools, platforms and registrys are already emerging to build, test, deploy and discover MCP servers. As the environmental system develops, AI applications will offer an easy interface to plug in new capabilities. Protocol -accepting teams will send products quickly with better integration stories. Public APIS and official MCP server companies can be part of the integration story. Late adoptions have to fight for compatibility.

Noah Schwartz is the head of the product Postman.

Daily Insights on Business Use Matters with Daily VB

If you want to impress your boss, the VB Daily covers you. We give you internal scope what companies are doing with Generative AI, from regulatory shifts to practical deployments, so that you can share insights for more and more ROIs.

Read our privacy policy

Thanks for subscribing. Check more VB Newsletter here.

There was a mistake.