I want to talk about something the MCP course glosses over entirely.
Not the technical spec. Not the primitives. Not how to wire up a server.
The politics.
The Industry Learned This the Hard Way
We did Betamax vs VHS. Better technology lost because one side got more studios on board and cheaper licensing. Betamax had superior picture quality. Didn’t matter.
We did Netscape vs Internet Explorer. Years of legal battles. Billions in antitrust fines. An entire generation of developers who would write a check to literally anyone except Microsoft. The browser wars didn’t produce a better internet. They produced trauma, fragmentation, and an industry-wide reflex that persists to this day: never let one company own the standard.
We did HD-DVD vs Blu-ray. We did every standards war you can name, and the lesson was always the same: the best technology doesn’t win. The one with the most political leverage does. And the winner always made enemies.
The lesson eventually produced USB-C — a neutral connector standard that exists specifically because no single company could be trusted to own it. It produced Kubernetes at the CNCF, because Google understood that if they kept it, Amazon and Microsoft would build competing systems rather than adopt something that handed Google an advantage.
Neutral governance isn’t idealism. It’s the mechanism the industry invented to let competitors collaborate without anyone losing face.
Anthropic remembered all of this when they designed MCP.
The Move
Anthropic built MCP in-house. A clean, well-designed protocol for connecting AI models to external tools and data sources. Think USB-C for AI — universal connector, any model, any tool, one standard.
Then they donated it to the Linux Foundation.
The moment they did, everything changed. Before the donation: Google adopting MCP meant Google helping Anthropic win. Microsoft adopting MCP meant handing a competitor a strategic advantage. Every major player had a reason to wait, fragment, or build a competing standard instead.
After the donation: Google adopting MCP meant being a good ecosystem citizen. Microsoft building MCP support meant following an open standard. The competitive risk evaporated. The path to yes was clear.
Within months, every major AI player was building MCP support. Tool vendors started writing connectors. The ecosystem consolidated around a single standard that, six months earlier, might have fractured into four incompatible ones.
Anthropic set the terms. Then made the terms belong to everyone.
That’s not naive. That’s sophisticated. That’s the Betamax lesson applied correctly, probably for the first time in AI.
What This Means If You’re Building on MCP
A few practical things worth knowing before you write a single line of code:
MCP is not going away. Independent governance means it’s not vulnerable to a single vendor’s business decisions. The spec lives at modelcontextprotocol.io. That’s a meaningful durability guarantee for anyone building on top of it.
The Python SDK is separate from the Anthropic SDK. mcp on PyPI and anthropic on PyPI are two different packages with two different purposes. One builds MCP servers and clients. One calls Claude’s API. You’ll likely use both. They’re not the same thing, and the docs live in different places. Worth knowing before you spend twenty minutes looking in the wrong place.
The missing install step nobody mentioned. The course tells you to run mcp dev mcp_server.py without explaining where the mcp command comes from. It comes from pip install mcp[cli]. One line. Not in the materials. You’re welcome.
The ecosystem is moving fast. Check whether an MCP server already exists for whatever tool you need before building a custom integration. The inventory is growing weekly. Odds are improving every month that someone already did the work.
The Lesson
Anthropic didn’t win MCP because they’re the smartest or the best-funded.
They won because they made winning uncontroversial.
The technical spec was good. The political move was better. And fifty years of standards wars is what made it possible to recognize the right move when it mattered.
The MCP courses at Anthropic Academy cover the technical spec thoroughly — free, certified, worth your time at anthropic.skilljar.com. The political subtext, you get here.
Next up: A 2006 technology nobody used for 15 years is now the backbone of every AI streaming response. Including the one delivering Claude’s replies word by word right now.