Anthropic's Model Context Protocol crossed 97 million installs on March 25, 2026 — a figure that represents the fastest adoption curve for any developer infrastructure standard in AI history. For reference: Kubernetes, the container orchestration system that now underlies most cloud infrastructure, took nearly four years to reach comparable enterprise deployment density.
MCP was introduced in late 2024. In roughly 16 months, it has gone from an internal experiment to the de facto integration layer for AI agents. Every major AI vendor — OpenAI, Google DeepMind, Microsoft, AWS, and Cloudflare — has integrated support, with formal backing codified through the Linux Foundation's newly formed Agentic AI Foundation.
In practical terms, what MCP does is standardize how AI agents connect to external tools and data sources. Before MCP, each AI platform had its own proprietary approach to connecting to a database, a CRM, or a code repository — meaning every integration had to be rebuilt for every platform. MCP eliminates that friction. There are now over 5,800 community and enterprise MCP servers covering cloud providers, databases, productivity tools, e-commerce platforms, and developer environments.
The 97 million install figure is significant not just as a milestone but as a signal of where enterprise AI is heading. Agentic AI — systems that take actions rather than just generate text — requires exactly the kind of standardized integration layer MCP provides. The protocol competition, such as it was, appears to be settled.
For the professionals who will be building on, managing, or simply using AI-powered systems over the next several years, MCP is the kind of infrastructure story worth understanding now, before it becomes invisible infrastructure.
That install base is also a security surface: Cyera agreed in July to buy Oasis Security for about $1 billion, betting that non-human identities are the fastest-growing identity type in the enterprise.
Comments