The agentic internet is coming. AI companies are racing to build the digital rails first. (www.businessinsider.com)

Big Tech and startups are racing to build the “rails” of an agentic internet — a new era where autonomous AI agents act on users’ behalf to search, book, negotiate and pay without screens or clicks. Cantor Fitzgerald’s recent report highlights rapid progress in protocols and SDKs from Google, OpenAI, Anthropic and others that enable agents to discover, authenticate, transact and access backend systems. The stakes are high: whoever defines these standards could lock in developer ecosystems and capture large economic value much like platform leaders did in the web and mobile eras. Technically, this stack replaces many legacy primitives (HTML, cookies, OAuth, API SDKs) with inter-agent messaging, verifiable credentials and agent-native payment flows. Key pieces include Google’s Agent2Agent (A2A) for structured inter-agent communication, Anthropic’s Model Context Protocol (MCP) to connect agents to databases and pricing engines, OpenAI’s Agents SDK/Google’s Vertex ADK for agent frameworks, and Google’s AP2 payments protocol (with Intent/Cart Mandates) to handle authorization, trust and fraud. Even Cloudflare is proposing NET Dollar, a dollar-backed stablecoin for agentic transactions. For developers, businesses and regulators, this signals a paradigm shift: commerce becomes intent-driven and programmatic, raising interoperability, security and governance questions as the industry races to own the new digital rails.