🎧 Listen to this article

Narrated by Talon · The Noble House

Anthropic released the Model Context Protocol in November 2025. In December 2025, they donated it to the Agentic AI Foundation — a directed fund under the Linux Foundation, co-founded by Anthropic, Block, and OpenAI. By early 2026, MCP had been adopted as the de facto standard for AI tool integration across the industry. When Peter Steinberger joined OpenAI in February 2026, his local-first agent architecture was already built on MCP. The protocol he didn't build is the foundation his work runs on.

Standards are boring until you realize the entire digital economy runs on them.

🎙️ Listen: Audio version

The Pre-MCP Fragmentation Problem

Before MCP, every AI model-to-tool integration required custom code. You wanted to connect Claude to a database? Write a connector. Connect GPT-4 to your CRM? Write a different connector. Switch models? Rewrite the connectors. The integration matrix was model-count multiplied by tool-count. In a world with five AI models and twenty tools, that's a hundred custom integrations to maintain. The overhead was real and the maintenance burden grew with every new model or tool added to the ecosystem.

MCP solved this by providing a universal protocol layer: a standardized format for describing tools, passing context, and receiving results that any MCP-compliant model can use with any MCP-compliant tool. Write the tool integration once in MCP format; it works with every model that speaks MCP. The integration matrix collapses from M×T to M+T.

The USB Parallel Is Precise

USB was proposed in 1994 when every peripheral device had its own proprietary connector: printers used parallel ports, mice used PS/2, modems used serial, keyboards had their own format. Adding a device required knowing which port it used and whether your computer had one. USB proposed a single connector for all peripherals. The resistance was real — every peripheral maker had invested in their proprietary interface — but the efficiency gains were large enough that industry adoption followed.

The result: you can plug any USB device into any USB port and it works. The standard made the peripheral market hundreds of times larger by removing the interoperability tax.

MCP is playing the same role in AI. The Wikipedia MCP article is direct: "MCP is the universal standard protocol for connecting AI models to tools, data and applications" (Linux Foundation press release, December 9, 2025). Anthropic's announcement noted that MCP "remains a neutral, open standard" — the language of a company that understands that proprietary standards die and open standards win.

MCP Model Context Protocol USB analogy pre-standard fragmentation versus post-standard interoperability
Pre-MCP: every AI-to-tool integration required custom code. The integration matrix was model-count × tool-count. Post-MCP: write the integration once in MCP format; use it with any MCP-compliant model. The USB analogy is precise: same connector, any device, any computer.

Why the Linux Foundation Move Was the Decisive Strategic Decision

If Anthropic had kept MCP proprietary, it would have failed as a standard. Standards require broad adoption; broad adoption requires trust; trust requires neutrality. OpenAI and Google would never build on an Anthropic-owned protocol — it would give Anthropic ownership over how their models integrate with the world. The competitive incentive to build alternative, incompatible standards would have been overwhelming.

By donating MCP to the Agentic AI Foundation under the Linux Foundation — the same organization that governs Linux, Kubernetes, Node.js, and dozens of foundational open-source projects — Anthropic transferred ownership to a neutral steward. OpenAI's co-founding of the Agentic AI Foundation is the proof that the neutrality worked: OpenAI contributing to MCP's governance rather than building an alternative means the standard has cross-industry legitimacy.

This is the same move that made HTTP work: CERN published it freely in 1993. The internet Berners-Lee built wouldn't have worked if HTTP were a CERN proprietary protocol that every other organization had to license.

The Ecosystem Value Being Created

Every tool built to MCP specification works with every MCP-compliant model. Every model trained to speak MCP can use every MCP tool without custom integration. The ecosystem being created looks like the app store economy that emerged from mobile operating system standardization — but operating at the protocol layer rather than the platform layer.

The organizations positioned to capture value in an MCP-standardized ecosystem are those building either: (a) the most useful tool integrations in MCP format (the application layer), or (b) the most capable models that speak MCP well (the infrastructure layer). The protocol layer itself is public infrastructure. The value sits above and below the standard, not in the standard itself.

MCP ecosystem value creation tool integration market above protocol layer model capability layer below
MCP creates a protocol layer that sits between AI models and tools. Value concentrates at two points: best-in-class MCP tool integrations above the protocol, and most capable MCP-native AI models below it. The protocol itself is neutral infrastructure — the ecosystem it enables is where competition happens.

What to Watch in MCP's Development

The protocol is still maturing. Security specifications for MCP tool access (permissions, sandboxing, audit trails) are under active development and will determine whether MCP scales into enterprise contexts that require rigorous access control. The current open-source governance model will face stress tests as commercial interests in the tools and models built on MCP grow.

The historical precedent is mostly positive: HTTP, TCP/IP, and USB all survived commercial pressure on the underlying standards because the network effects of interoperability were large enough to discipline defection. The same logic likely applies to MCP — but "likely" is worth watching, not assuming.

Sources: Anthropic, "Donating the Model Context Protocol and establishing the Agentic AI Foundation," December 9, 2025; Linux Foundation, "Formation of the Agentic AI Foundation," December 9, 2025; MCP Blog, "MCP joins the Agentic AI Foundation," December 9, 2025; Wikipedia, "Model Context Protocol" article; CNBC, OpenClaw/Steinberger joining OpenAI coverage, February 15, 2026

Sources