When Anthropic published the Model Context Protocol in November 2024, it was framed as an open standard for connecting AI agents to external data and tools. Eighteen months later, it has become something else: its own category of venture capital investment, with dedicated startups, dedicated funding rounds, and adoption by every major AI provider in the world.

MCP has reached 97 million monthly SDK downloads. OpenAI, Google DeepMind, Microsoft, and AWS have all adopted it. Anthropic donated the protocol to the Linux Foundation’s Agentic AI Foundation, co-founded with OpenAI and Block, cementing it as neutral infrastructure rather than proprietary technology.

And now there is a cluster of well-funded startups whose entire reason for existing is MCP.

What MCP Actually Solves

The problem MCP addresses is deceptively simple to state and genuinely hard to solve in practice.

AI language models are powerful reasoning engines, but they are isolated systems. Without external connections, a model can only work with what is in its context window. To do anything useful in a real business environment, an AI agent needs to read from databases, write to systems of record, call APIs, access files, and interact with the tools that businesses actually run on.

Before MCP, every integration between an AI system and a business tool was a custom build. Every AI developer connecting their product to Salesforce, Notion, Jira, or a proprietary database had to build that connection from scratch, maintain it, and rebuild it every time either side changed.

MCP is a standardised protocol that eliminates that redundancy. An AI agent that speaks MCP can connect to any tool or data source that has an MCP server, without a custom integration for each pair. It is the same principle that made HTTP a foundation layer for the internet rather than a proprietary connection scheme.

Why This Created a Funding Category

The moment MCP became an industry standard rather than Anthropic’s internal protocol, the infrastructure question became urgent. Enterprises could not simply run MCP servers on developer laptops. They needed hosted infrastructure, security controls, identity management, monitoring, and scalability.

That gap created the opening for a new category of company. Three have raised notable rounds in the past twelve months:

Manufact (formerly mcp-use), a Y Combinator Summer 2025 company, raised $6.3 million in February 2026, led by Peak XV with Liquid 2 Ventures, Ritual Capital, Pioneer Fund, and YCombinator participating. Manufact builds an open-source MCP SDK and a cloud hosting layer that lets developers deploy and scale MCP servers without managing their own infrastructure.

Runlayer raised $11 million in November 2025, led by Khosla Ventures via Keith Rabois and Felicis Ventures. Runlayer addresses the security problem: as enterprises deploy AI agents that connect to real business systems via MCP, those connections need authentication, authorisation, audit trails, and threat monitoring. Runlayer builds that security layer.

Alpic raised $6 million in pre-seed funding in September 2025, led by Partech. Alpic is building what it describes as the first MCP-native cloud platform: infrastructure designed from the ground up for deploying and scaling MCP servers, rather than adapting existing cloud infrastructure to a new protocol’s requirements.

Three funded startups in one protocol’s ecosystem, with Khosla and Partech among the investors, signals that serious investors see MCP as a durable infrastructure layer, not a passing trend.

The Agentic AI Foundation

In early 2026, Anthropic announced it was donating MCP to the Linux Foundation and co-founding the Agentic AI Foundation alongside OpenAI and Block. That move was significant for several reasons.

By placing MCP under neutral governance, Anthropic removed the primary objection that would prevent competitors from fully embracing it. A protocol controlled by a competitor creates strategic risk. A protocol controlled by a neutral foundation can become a true standard.

The co-founding of that foundation with OpenAI, Anthropic’s most direct rival, was a statement that both companies see MCP adoption across the industry as a positive outcome. When competitors cooperate on infrastructure, it typically means both believe the infrastructure layer will be valuable regardless of which AI models sit above it.

For businesses, that competitive dynamic matters because it reduces the risk of building on MCP. A protocol with this level of cross-industry adoption is not going to disappear or become a dead end. It is becoming load-bearing infrastructure.

What This Means for Businesses

If your business is deploying, evaluating, or planning to deploy AI agents in the next twelve months, MCP is the connectivity layer you will likely encounter.

The practical implication is that AI agents built on MCP-compatible infrastructure can connect to more of your systems with less custom engineering effort. The more MCP-compatible tools exist in your stack, the lower the integration cost for new AI deployments.

The secondary implication is security. As AI agents connect to real business systems at scale, the access controls around those connections matter enormously. The fact that dedicated MCP security infrastructure already exists and is well-funded suggests the industry is taking this seriously. Your security posture for AI agent deployments should include MCP access governance, not just model-level controls.

97 million monthly SDK downloads is not a protocol in early adoption. It is infrastructure that is already being built on, at scale, across the industry.

---

Enterprise DNA helps businesses design AI agent architectures that connect to the tools and data you already have. Talk to us about what an enterprise MCP deployment looks like for your organisation.