Source: a16z
Compiled by: Felix, PANews
The era of Agentic Commerce has arrived.
OpenAI’s AI Agent Commerce Protocol (ACP) and Google’s Universal Commerce Protocol (UCP) promise to enable checkout functionality within ChatGPT and Gemini. Soon, hundreds of millions of consumers worldwide will discover better products, merchants will see improved conversion rates, and platforms will earn a 5–10% cut.
However, ChatGPT’s checkout feature is merely a gradual improvement. It won’t reshape society the way the internet did in the early 21st century—but open agent commerce will.
We must go back to the 1990s to understand why.
At the time, there were two competing versions of the "internet."
AOL version: unified pricing, email, weather forecasts, additional moderation content, and eventually the entire Time Warner content library.
Open protocols: HTTP, DNS, HTML, and a browser called Mosaic.
Compared to AOL’s version, Mosaic seemed absurdly primitive. There were so few websites on Mosaic that search functionality was unnecessary—a simple alphabetical index was sufficient. Eight years later, AOL reached an equal merger agreement with Time Warner worth $350 billion. The market had made its choice: curated content appeared to be the future.
But not long after, Mosaic and open protocols emerged victorious, and human civilization officially entered the digital age. Why? Imagine what would have happened if closed ecosystems had prevailed instead.
In 2004, Zuckerberg wanted to launch Facebook and needed a distribution agreement with AOL. Two Stanford students wanted to build a web index and needed permission from CompuServe. Someone wanted to sell books online from their garage and had to submit a proposal to MSN’s content team.
They would say: “Go back to school, kids.” None of these things would ever happen. The entire digital economy we take for granted simply wouldn’t exist.
Open protocols mean there are no "gatekeepers." Anyone with a server and a domain can access the entire internet. Innovation constantly emerges from the edges, while the center struggles to keep up—ultimately giving rise to one of the largest wealth-creation events in human history. This is a fundamental principle of capitalism: innovation originates at the edges.
Back in 1997: Tim Berners-Lee, Marc Andreessen, and others were developing protocols and browsers. At the time, running a server cost hundreds of thousands of dollars. It was unclear why a content server should respond to requests from strangers—doing so was expensive and offered no known economic incentive.
They created a message status code called "402," which servers could send to users: "Pay a small fee to access this content." But at the time, there was no practical way to make digital payments. PayPal had not yet been invented, and credit card fixed fees of dozens of cents were far too high for one-cent microtransactions.
Yet, the internet still took off.
Google found a unique business model for the internet: advertising. In traditional media, the primary economic relationship exists between content creators and consumers. Google adopted the economic model of broadcasting by introducing a third party: advertisers, who pay for the relationship between content creators and consumers.
This is very clever. Today, producers can monetize audience attention without needing to establish any prior relationship with consumers. Google sits securely in the middle of this flow of money, between advertisers and content creators, taking any percentage it desires.
As a result, the demand for microtransactions was avoided. Open-source software began to take off, the cloud computing revolution erupted, and hosting costs plummeted by a factor of a hundred. Google became the largest advocate of a free and open internet. The more consumers searched, the more money Google made. Consequently, they invested hundreds of billions of dollars to make the internet fast, cheap, and ubiquitous.
Then came the 2010s, and everything seemed unchanged.
Low interest rates, slow technological advancement, and growing closed ecosystems.
In 2022, ChatGPT launched, and the world was about to change once again. Large language models (LLMs) can do more than just provide results—they can generate and compile many such results into a clear, concise summary, often without ever touching the original content.
By the time GPT-4 emerged, it was clear: agents were the next frontier—LLMs excel at using computers like humans, but at lower cost and higher efficiency.
And so, the economic landscape of the internet has changed.
From 1997 to 2024, the core business model was "distraction marketing." Humans browsing web pages are easily distracted by advertisements, which advertisers exploit to profit from their limited attention. Lower-level agents/agents, however, are not distracted.
There lies a curious irony: advertising created the free and open internet, and the internet, in turn, gave rise to massive datasets worth ten trillion tokens, which led to the development of LLMs, ultimately contributing to the decline of advertising.
Since the release of GPT-4, Stack Overflow traffic has dropped by 75%, and tech news traffic has fallen by 60%. Tech users are early adopters, but this trend will eventually sweep through all information online.
The checkout feature in ChatGPT is not important. The internet has become the public square of civilization, and traditional economic contracts are already outdated.
On the internet, a few niche areas have successfully resisted Google’s “erosion,” serving as “walled gardens” with truly unique content: Facebook, TikTok, and LinkedIn. These areas have been able to resist automated web crawlers thanks to thousands of highly paid engineers working around the clock.
But the walls of the walled garden have been breached. Agents with computer literacy can perfectly simulate the traffic of real human users. Charlatans will sell various “magic bullets” over the next decade, and capital from Sand Hill Road (PAnews note: Silicon Valley’s venture capital hub) will follow suit. In reality, there are no magic bullets. The fortress walls have been destroyed by fighter jets.
What’s next?
Open Agent Commerce.
The checkout feature in ChatGPT is like AOL in the era of agent-based commerce—a curated directory, a walled garden with a superior user experience. To sell through it, businesses need months of business development, stringent legal documentation, a detailed five-year plan, substantial revenue, a strong user base, and a compelling story that will please shareholders when featured on the front page of The New York Times.
Open agent commerce is like today’s HTTP. It’s a simple set of protocols that allows agents to pay for anything they need—data, cloud hosting, communication, and many things we haven’t yet imagined.
Coinbase's x402 and Tempo, developed in partnership with Stripe, are the two leading MPPs. After 28 years since the invention of the "402" status code, we finally have a viable implementation. Stablecoin transactions on modern blockchains cost less than a cent, solving the fixed-fee problem that caused micropayments to fail in 1997.
An agent that can only purchase from pre-approved merchants is like an employee with a corporate credit card that can only be used at three suppliers. An agent with an open protocol is like an entrepreneur with a bank account.
There is no BD, no whitelist, only simple, permissionless standards.
These protocols focus on only two things:
How do I make a payment?
Merchant: "How can I ensure the agent has been paid?"
LLMs excel at calling tools they have never seen before. Starting with Claude 4.5+ and Codex 5.2+ models, agents can discover APIs, read their schemas, and use them correctly without prior training.
The current discussion focuses primarily on "skills." You can think of them as natural language programs that can be combined like building blocks. A founder without a technical background can write a Slack message and have it execute like software:
Order a pizza from a highly-rated local pizzeria and track the delivery status every 10 minutes.
Turn on the porch light when the delivery person is 5 minutes away.
Tip the delivery person $5 if the order is delivered within 30 minutes.
No coding required, no computer science degree needed. The agent reads your intent, instantly writes a native computer program, executes it, and then discards it. Programming no longer needs to be a discipline—mastering your native human language is enough.
Skills are indeed effective. But they are merely a transitional solution—the first approach we thought of after discovering that agents could invoke unfamiliar tools. They require someone to write, publish, perform security checks, and update them. Additionally, agents need to preload them in advance. This is cumbersome.
Discussions about skills obscure a deeper breakthrough: agents can combine capabilities in unprecedented ways.
Buying pizza is just a simple example. Here’s a more realistic case: an agent managing a small business’s supply chain notices that packaging supplier prices have increased by 15% due to tariffs. It identifies three local alternatives, requests samples from each, negotiates bulk pricing, and completes the switch—all before the business owner finishes their morning workout.
No API partnership, procurement team, or bidding process required—just an agent with a balance in an account and using an open protocol.
Discover
The agent can make payments and combine capabilities, but cannot yet locate the required content.
The remaining question is “discovery.” For agents: “How can I find what I want to buy?” For merchants: “How can I introduce my services to agents?”
AgentCash has emerged. It is a unified account that grants access to all APIs on the internet. Even when agents are blocked, it can access thousands of APIs and continue operating at minimal cost.
The key is that AgentCash integrates payment and merchant discovery functionality. Merchants can register their servers on x402scan.com or mppscan.com and immediately make their services visible to all 2,000+ AgentCash agents.
In 1997, the internet had no business model, and no one understood why servers should communicate with strangers. Open protocols and a strategy called "advertising" elegantly solved this problem, ushering civilization into the digital age. By 2026, this elegant strategy is fading away, replaced by open protocols and a 28-year-old status code.
Welcome to the era of open agent commerce.
Read more: Understanding x402 and MPP: Two Paths for Agent Payments