MCP: how an open protocol became the USB-C of AI agents
In very little time it went from an internal Anthropic experiment to a standard adopted by OpenAI, Google and Microsoft. Why it matters for anyone connecting an agent to their business.
Until recently, connecting an AI agent to a company's tools was artisanal work: each model, against each system, with a hand-built integration. The classic M × N problem. Then Anthropic published the Model Context Protocol (MCP) — a standard way for any agent to talk to any tool. In very little time it stopped being a single-vendor proposal and became common infrastructure.
Why it won
MCP solved the right problem in the simplest way: instead of N integrations per model, you define a tool once and any compatible agent uses it. It's open, it's simple, and it arrived at the right time. OpenAI adopted it early across ChatGPT and its Agents SDK; Microsoft integrated it into Windows, Azure and Microsoft 365; Google confirmed support in Gemini. When your competitors adopt your standard, it stops being your standard — it becomes the market's.
From spec to public infrastructure
Later, Anthropic donated MCP to the newly formed Agentic AI Foundation, a fund under the Linux Foundation backed by Google, Microsoft, AWS and OpenAI. The signal is clear: MCP no longer belongs to one company, it belongs to the industry — the way HTTP or USB-C did.
"MCP became for agents what USB-C was for devices: a connector that just works, regardless of brand." — The New Stack, Why the Model Context Protocol Won
What changes for your business
For anyone about to put an agent to work, this matters more than it seems. An agent built on open standards connects to your CRM, your calendar or your inventory without bespoke integrations that break, and doesn't lock you to a single AI vendor: if it makes sense to switch models tomorrow, the agent keeps running. At InnovaBlack we build on open standards for exactly this reason — portability and no lock-in, because the agent is yours, not the vendor's.