I was angry.
While building an agent-related service, I wanted to connect a few Korean services, so I requested API access from a major Korean company. This company had already put its own service on ChatGPT. It was using OpenAI's window, yet replied this way to an outside request to use its own window.
"We're currently in discussions with another platform company, so it's difficult to open this up to external parties."
They freely walk through other people's doors while keeping their own firmly locked. This contradiction is the current state of Korea's tech ecosystem.
Developers and startup founders around me have had similar experiences. When they request API integrations from Korean companies, the answers are usually "Business review required," "Under internal review," or silence. This isn't a problem with individual companies. It's a structural closedness embedded throughout Korea's tech industry.
The numbers show it. Korea has 2.66 million developers, putting it near the top globally. The problem isn't the number of developers. It's that they have no doors to walk through. More than 10,000 services are already registered in the MCP ecosystem, but only a tiny number connect with Korean services. While global companies like Zapier, Expedia, and Canva have established themselves as essential tools in the GPT Store, the names of major Korean companies are hard to find.
Look at Kakao. You can integrate with it relatively freely for basic app development, but the moment you move into areas that actually make money, such as AlimTalk or business channels, a review is required. The criteria for that review aren't public. More than a few developers must have run into this wall while quickly building something over a weekend.
Korea's open banking system required years of discussion before it was fully implemented in 2019. In the meantime, Plaid became the standard for connecting financial data in the United States. Ironically, Plaid was born because banks wouldn't open their doors. With no window available, it climbed over the wall, and the US turned that workaround into a foundation for innovation. In Korea, countless fintech startups withered away in a maze of regulations during that same period.
Just because something has "open" in its name doesn't mean it is open. Korea's MyData policy is like that. The name is open, but participating requires a license, and getting a license requires meeting capital and staffing requirements and passing security certifications. It's not a structure that a startup founded by three people can access. The EU, by contrast, legally forced banks to open their data, and companies specializing in financial connectivity, such as Tink and TrueLayer, grew in that environment.
In the end, whether regulation kills or sustains an ecosystem depends on how the word "open" is defined.
The history of US tech is a history of connection. Stripe made it possible to add payments with a few lines of code. Twilio abstracted phone calls and text messages into connection tools and became a company with more than $5 billion in annual revenue. Once Salesforce opened its platform, the partner ecosystem is estimated to exceed $2 trillion by 2028. More than 10,000 apps are built on Shopify. These companies have one thing in common. They understood that the value of their platforms would grow through outside developers, and they acted on that belief.
Korea's mainstream companies define sharing the value of their platforms with the outside world as a risk. There are many reasonable explanations. They worry that competitors will use it, that data will leak, or that they will lose control. So, under the name of "strategic partnerships," the representatives of large companies take photos together beneath banners and connect only with one another, while locking the door to everyone else.
This logic worked in the web era. Naver dominated the Korean market through search, and Kakao through messaging. They just had to build their walls high and do business inside them. Walls of similar height went up across various other sectors as well.
In the agent era, the rules of the game change. Without an API as a connection point, no outside action can happen. Newly born entities don't look at screens like people do. They read the world only through the radar of APIs. No matter how beautiful the interface or how good the service, if it doesn't send out an API signal, it looks like a massive wall. If the walls of the web era were moats that kept competitors out, the walls of the agent era are prisons. Spaces where everything works perfectly inside, but that are completely invisible from the outside.
Someone tells an AI agent, "Book me a plane ticket." The agent explores connected services. US flight-booking services and Southeast Asian commerce apps respond one after another. Almost every Korean service has no window at all. So global users don't even realize that Korean services are missing.
Paradoxically, opening the door creates a stronger and wealthier castle. When Stripe opened its payment window, hundreds of thousands of developers built their own services on top of it, and those developers soon became the shield protecting Stripe. If they switched to another payment tool, hundreds of thousands of services would collapse with them. You don't protect yourself by closing. You protect yourself by opening.
If this were a matter of technical capability, the answer would be clear. Hire more engineers and invest more. But this is a matter of culture and philosophy. We need a shift in perception, from seeing connection as a "risk" to seeing it as an "opportunity." It's an old habit of distrust toward the outside world. As a result, Korean startups build on OpenAI, AWS, and Google Cloud instead of Korean platforms. Korean platforms are left alone without an ecosystem, while Korean startups lose the foundation for building tools optimized for Korean people.
The foundational infrastructure of the agent economy is still being built rapidly. x402 and ERC-8004 are that infrastructure.
x402 is the "agent's wallet," and ERC-8004 is the "agent's passport." x402 is a standard that allows AI agents to make payments on their own, without human intervention. Until now, even when AI needed a service, a human had to take out a credit card. This protocol, launched by Coinbase in May 2025, became established infrastructure when Coinbase founded the x402 Foundation together with Cloudflare in September of the same year. It processed more than 100 million transactions within six months of its launch and was adopted by Google, Visa, and Vercel.
ERC-8004 gives hundreds of millions of agents blockchain-based identities and reputation records so they can trust one another. Engineers from MetaMask, the Ethereum Foundation, Google, and Coinbase drafted it in early August 2025, and it was deployed on the Ethereum mainnet in January 2026. Once the wallet and passport work together, an economic ecosystem is complete in which agents verify one another, transact, and build reputations.
It is difficult to find any major Korean conglomerates that support x402 or participate in ERC-8004. While companies overseas build the payment networks and identity infrastructure of the agent economy together, it is not even clear whether Korean companies recognize this trend.
Even now, the economic map of the agent era is being drawn. Keeping the door closed doesn't block competitors. It erases you from the map. They won't wait at a closed door. They won't even knock. They simply move on.
