A new Internet age is shaping, not a screen, click, or even a swipe. AI agents, autonomous software programs acting on your behalf, get transactions, book flights, manage workflows, and even pay for things without lifting your fingers.
The architecture that drives this age of agent is currently being built, with large tech companies competing to establish early technological footholds.
According to Cantor Fitzgerald's latest BYTES report, over the past six months, we have seen rapid advances in the underlying protocols that have enabled these AI agents to work across platforms. And like the innovations of the railways and the previous frontiers, race is now defined The rails of this new digital economy and the tech giants, including Google, Openai and humanity, are laying the foundations.
Why is this important? As Cantor pointed out, anyone who sets the standard wins in the long run. Just as Google and Apple locked up the developer ecosystem in the web and mobile era, pioneers of today's protocols could rule the future of agents.
“Monitoring who is building, maintaining and supporting these frameworks will help us evaluate competitive moats over the long term,” writes Kanter analysts Deepak Mattivanan and Jack Halpert. “So companies pioneering the standards have historically benefited from ecosystem lock-in and have acquired a large portion of economics over the long term.”
Goodbye App, Hello Protocol
Cantor Fitzgerald Research Chart Canter Fitzgerald
The web was constructed with HTML pages and navigated through hyperlinks and HTTP protocols. Cookies are used to store session state on web pages, and security is maintained through an SSL certificate. SQL or NOSQL queries enable database access in Business Logic and payment processing is performed through the API at an external gateway.
In the mobile age, apps are built using native OS programming languages such as Swift, Java, and Kotlin. These apps communicate through mobile deep links. Apple ID and Android ID are used to store session state on your smartphone. Authorization and certification are carried out using OAUTH 2.0 tokens originally released by Twitter and widely adopted by Google and others. Payment processing is processed using SDKs such as Stripe's mobile SDK and mobile wallets such as Apple Pay API using NFC communication.
The new agent era is built on a fundamentally different technology stack. Instead of user-driven apps, agents handle intent through frameworks such as Openai's Agent SDK and Google's Vertex ADK. They don't click, they communicate and do it using protocols.
Google's Agent2Agent (A2A) has emerged as Lingua Franca, an agent-to-agent communication. Launched in April, A2A allows for a secure, structured exchange between AI agents representing users and other agents representing merchants. This is a foundation layer similar to HTTP for the web and deep links for mobile.
Meanwhile, Anthropic's Model Context Protocol (MCP) connects agents to backend systems such as databases, pricing engines, and workflows, replacing patchwork and power legacy commerce in APIs online.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Recently, Google has deployed AP2, an agent-native payment protocol. It aims to tackle trust, approval, fraud and help users issue “intentional obligations” and “cartmandate” that allow users to complete transactions on behalf of AI agents. Google has worked with over 60 companies on this, including PayPal, American Express and Etsy.
On Thursday, CloudFlare said it plans to introduce Net Dollar, a US dollar-backed stablecoin designed to enable secure transactions for Agent Web.
Commerce has been rewritten
Shifts are more than technical. Rethink how transactions are carried out. In the traditional web model, users navigate through hyperlinks, authenticate with cookies, and pay via APIs. In the age of agents, users simply express their intentions, such as “booking flights to Tokyo for under $1,000”, while AI agents do the rest, from database queries to payments.
Cantor analysts shared an illustration of the report showing this end-to-end flow. The user agent calls the merchant agent via A2A, queries the inventory via MCP, authenticates via verifiable credentials, and completes the transaction on AP2.
Flowcharts created by Cantor Fitzgerald Analysts Canter Fitzgerald
It was early on, but adoption was already accelerating, analysts said. For developers, businesses and investors, this could be a paradigm shift in how digital commerce works in the AI era.
Welcome to Agent Internet. The competition to own it is on.
Sign up for BI's Tech Memo Newsletter here. Please contact me by email abarr@businessinsider.com.
