Anthropic’s Model Context Protocol crossed 97 million installs in March 2026. Every major AI provider — OpenAI, Google DeepMind, Microsoft, Meta — now ships MCP-compatible tooling. The protocol has moved from an experimental Anthropic specification to the foundational infrastructure layer of the agentic AI era in under eighteen months. The speed of that transition is without precedent in the history of AI infrastructure standards.
What MCP Does and Why It Spread
The Model Context Protocol defines a standard way for AI models to interact with external tools, data sources, and services. Before MCP, every AI agent framework had its own integration layer — a fragmented landscape where connecting a model to a database, a code executor, or an enterprise API required custom connectors for each combination of model and tool.
MCP replaced that fragmentation with a single open interface. A tool built to the MCP specification works with any MCP-compatible model, and any MCP-compatible model can use any conforming tool without additional glue code. This is roughly analogous to what USB did for device connectivity or what HTTP did for data transfer on the web — a common interface that enables a network effect.
The numbers reflect this: 97 million monthly SDK downloads, more than 10,000 public MCP servers, and adoption from every major AI lab. For comparison, Kubernetes — arguably the most successful infrastructure standard of the last decade — took nearly four years to reach comparable enterprise deployment density.
Governance and the Linux Foundation
In December 2025, Anthropic donated the protocol to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, co-founded by Anthropic, Block, and OpenAI. This move was as strategically important as the protocol’s technical design.
By moving MCP into neutral governance, Anthropic removed the primary objection that would have prevented competitors from adopting it at scale: that they were building on infrastructure owned and controlled by a direct rival. Open governance converted MCP from an Anthropic product into shared industry infrastructure — and cleared the path for the cross-vendor adoption that produced the 97 million install figure.
The governance structure also imposes coordination costs on any competitor that might try to fork the protocol or introduce an incompatible alternative. With OpenAI, Google, and Microsoft all invested in MCP’s success, the protocol has the kind of institutional lock-in that tends to persist.
What It Means for the Agentic AI Ecosystem
The standardization of agent-tool communication is one of the preconditions for enterprise AI deployment at scale. CIOs and engineering leaders have been reluctant to commit to agentic architectures while the integration landscape remained fragmented — the risk of betting on the wrong tooling protocol was real. MCP’s emergence as a clear standard removes that risk.
The practical effect is already visible at NVIDIA’s GPU Technology Conference in April 2026, where agentic AI frameworks drew the largest attendance and Fortune 500 companies announced production deployments across manufacturing, logistics, and finance. Many of those production deployments are built on MCP.
Anthropic designed the protocol and donated it to a foundation, but the company that benefits most from MCP’s success is not necessarily Anthropic. The standard benefits whoever can build the most useful tools and the most capable models on top of it. That competitive dynamic is now fully open.
