Written by: Sam Ragsdale

Translated by: Chopper, Foresight News

The era of agent commerce has already arrived. The ACP and UCP protocols promise to enable one-click checkout within ChatGPT and Gemini. Soon, hundreds of millions of consumers worldwide will easily find better products, merchants will see significant increases in conversion rates, and platforms will also enjoy a 5%-10% share.

However, the built-in checkout in ChatGPT is just a incremental improvement; it cannot reshape society like the early 21st-century internet did—only open agent commerce can achieve that.

Why Walled Gardens Were Ultimately Overturned by Open Protocols

To understand the reason, we must trace back to the 1990s.

At that time, there were two distinctly different, competing "forms of the internet."

The AOL model: unified monthly fees, integrated email, weather, and other curated content, eventually encompassing the entire Time Warner copyright library.

The open protocol model: HTTP, DNS, HTML, and a browser called Mosaic.

Compared to the former, Mosaic seemed absurd at the time: there were very few websites, and there was no need to search; alphabetical indexes sufficed. Eight years later, AOL merged with Time Warner for $350 billion, and the market once believed that curated content was the future.

But it didn't take long for Mosaic and the open protocols to prevail, officially ushering humanity into the digital age.

Why? Let's suppose that the closed walled gardens had won instead.

In 2004, Mark Zuckerberg wanted to start Facebook, but he had to negotiate distribution with AOL. Two Stanford students wanted to build a web index and needed permission from CompuServe. If someone wanted to sell books online from their garage, they would have to pitch to Microsoft’s MSN content team.

They would only be discouraged: "Go back to school, young people." If that were the case, the digital economy we now take for granted would not exist.

The essence of open protocols lies in the absence of gatekeepers. Anyone with a server and a domain name can reach the entire web. Innovation is born at the margins, the center struggles to catch up, ultimately creating one of the largest waves of wealth generation in human history. This is the underlying logic of capitalism: disruption always comes from the margins.

Back to 1997. Tim Berners-Lee and Marc Andreessen were researching related protocols and browsers. At that time, setting up a server could cost hundreds of thousands of dollars, and no one understood why content servers should respond to requests from strangers: it was expensive, with no clear profit.

They designed a status code called "402," allowing servers to return a prompt to users: "Pay to access this content." But digital payments were completely unfeasible at that time: PayPal had not yet been born, and fixed fees for credit card transactions were often dozens of cents, far too high for $0.01 microtransactions.

Even so, the internet still rose.

Google found an alternative business model: advertising. In traditional media, content creators establish direct economic relationships with users. Google, relying on the logic of broadcasting economics, introduced third-party advertisers, who pay to maintain content supply and user access.

Ingenious. Creators do not need to build user relationships; they can monetize solely based on attention. Google occupies the flow of funds, wedged between advertisers and content creators, easily collecting commissions.

The need for micropayments was shelved. Open-source software took off, the cloud computing revolution erupted, and the costs of server hosting plummeted by a hundredfold. Google became the staunchest advocate of the open free internet: the more users search, the more Google profits. To achieve this, it invested hundreds of billions of dollars to make the web faster, cheaper, and ubiquitous.

AI Agents End Traditional Advertising and Walled Defenses

Fast forward to the 2010s, the industry stagnated.

Interest rates continued to drop, capital grew conservative, and the underlying innovation of the internet lost its former luster, as major walled gardens gathered users and built up strength.

In 2022, ChatGPT went live, and the world was transformed. Large language models not only output results but also integrate multi-source information to generate concise summaries, often without needing to directly fetch native content.

By the time GPT-4 arrived, the trend was already clear: agents would become the next core. They operate computers as skillfully as humans, at lower costs and higher efficiency.

At this moment, the fundamental economic logic of the internet has been completely rewritten.

From 1997 to 2024, the core of internet profitability was attention: humans are easily distracted by ads while browsing, and platforms monetize fragmented attention. But large language models and agents are not easily disturbed.

A deeply ironic cycle. Advertising gave birth to a free and open internet, massive online corpus trained large language models, which ultimately turned around to end advertising.

After GPT-4 launched, Stack Overflow's traffic plummeted by 75%, and technology news sites saw a 60% drop in visitors. Tech users, who were early adopters, will see the transformation eventually sweep through all information scenes online.

The built-in checkout in ChatGPT is insignificant. The internet was meant to be a plaza for human civilization, and the old commercial contracts have already expired.

There are still a few corners of the internet relying on differentiated content to resist Google's scraping, namely the classic walled gardens: Facebook, TikTok, LinkedIn. Thousands of highly paid engineers work around the clock to shield against machine crawlers.

But now, the walled defenses have collapsed. AI agents can perfectly mimic human online operations, fooling all protective mechanisms. In the next decade, countless schemes claiming to crack this situation will emerge, and venture capital firms will keep pouring money, but there will be no truly effective solutions. Just like after the advent of fighter jets, no matter how solid the ground fortress, it will inevitably be phased out by the times.

The Era of Open Agents

What will happen next? The answer is open agent commerce.

The built-in checkout in ChatGPT is akin to AOL in the era of agent commerce: curated catalogs, closed loops, optimized experiences. For merchants to sell their goods through it, they need months of business development, rigorous legal documentation, concrete five-year plans, substantial revenue, a large user base, and a story that can make it to the front page of The New York Times and thrill shareholders.

Open agent commerce is akin to the current HTTP protocol. A minimal universal standard allows agents to pay on demand for everything: data, cloud hosting, communication services, and an endless array of yet-to-be-born new scenarios.

Two pioneers have already landed, with Coinbase launching the x402 protocol and Tempo collaborating with Stripe to launch the mpp protocol. Twenty-eight years after the "402" status code was born, a feasible landing plan has finally emerged. Modern blockchain stablecoins have fixed transaction fees of less than 1 cent, perfectly overcoming the cost issues that stifled micropayments in 1997.

Agents that can only purchase from a pre-approved whitelist of merchants are like ordinary employees with limited corporate card access; those connecting to open protocols are akin to entrepreneurs with free bank accounts.

No business negotiations, no list reviews—only simple standards that require no permission.

These protocols mainly focus on two things:

• Agent side: How to make payments easily?
• Merchant side: How to verify payment receipt?

Large language models excel at invoking tools they have never seen before. Starting with the Claude 4.5+ and Codex 5.2+ models, agents can discover APIs, read their patterns, and use them correctly without prior training.

Current discussions are centered on skills. They are essentially natural language programmable modules that can be freely combined and assembled. A non-technical entrepreneur can automatically execute a program from a casual piece of text:

• Purchase pizza from a well-rated pizzeria nearby and check the delivery status every 10 minutes.
• Turn on the porch light when the driver is 5 minutes away.
• If it arrives within 30 minutes, give the driver a $5 tip.

No coding, no programming background required. Agents understand intent, generate native programs in real-time, and destroy them once executed. Programming is no longer an essential professional skill; mastering natural language suffices.

Skills are indeed effective but are merely a transitional product—the first intuitive landing form since humans discovered that agents can invoke unfamiliar tools. It requires dedicated personnel to write, publish, verify security, and iteratively update, and needs to be pre-installed, making it cumbersome and inefficient.

The skills craze masks deeper disruption: agents can combine various capabilities in ways never seen before.

Pizza is just a simple example; real business scenarios are far more transformative. Supply chain management agents for small businesses can detect that tariffs have caused packaging suppliers to raise prices by 15%, autonomously explore three local alternative vendors, request samples, negotiate bulk pricing, and switch partners with one click. All this can be done before the boss's morning workout.

No need for API integration, procurement teams, or bidding processes—just a balance account and open protocols.

Agents can make payments and create, but currently, they still struggle to find what they need.

What remains is exploration; for brokers: "How do I find what to buy?" and for merchants: "How do I introduce my services to brokers?"

Currently, the industry has built a universal search registry ecosystem; service providers just need to complete node registration at x402scan.com or mppscan.com to access the network of agents compatible with open protocols, achieving standardized supply-demand connections and automatic micropayment settlements.

In 1997, the internet lacked business models, and no one understood why servers should respond to strangers. Open protocols and advertising cleverly broke the deadlock, ushering civilization into the digital age.

In 2026, the makeshift plan of advertising will come to an end. Open protocols and a state code long sealed for 28 years will take its place.

Welcome to the new era of open agent commerce.