Model Context Protocol (MCP), first defined by Anthropic, is surfing a wave of enthusiasm right now. MCP has, incredibly quickly, become a de-facto standard for integration with LLMs. OpenAI’s announcement of support for MCP in the last week of March 2025 essentially confirms this position. New announcements of MCP support are emerging on an almost hourly basis.

Away from the wave of MCP servers and integrations, the bigger question that arises is how to interpret Anthropic’s underlying strategy with MCP, and what are the wider implications for the industry. We briefly look at this from two angles, consumer vs enterprise positioning and platforms.

Consumer vs Enterprise Positioning

Anthropic is not positioned as a consumer brand. Everything in their actions and intent thus far has been positioned towards enterprises or power users. Yes, they support consumers, but not as the first port of call. ChatGPT and by extension OpenAI is, as Ben Thompson at Stratechery often notes, an accidental consumer brand at scale. That is not to say that OpenAI is not focused on enterprise, they absolutely are, but their GTM is distinctly different from that of Anthropic.

While Anthropic, OpenAI and other foundation model providers have similar problem sets to deal with, this focus on enterprise and power users means that Anthropic have a much more urgent need to bring external context into the LLM and allow for additional integration points. Hence MCP.

From a Platform Layer to the Platform Core

Enterprises, and software vendors in particular, currently view LLMS as being a layer of a platform they leverage, providing input to and consuming its output. The bottom of the implementation pyramid one might say.

LLMs as a Platform Layer

However, Anthropic, and probably all new LLM providers, see the LLM very much at the core of what happens next. They provide analysis, decision and orchestration layers. Which in turn increases usage and ultimately drives revenue. The tools using the LLM are there to serve it, provide context and execute its decisions. Not the other way round.

LLMs as a Platform Core

But how do you manage such a feat, you can’t just disintermediate the tools – you need the tools to disintermediate themselves. That, at its core, is the goal of MCP and in my opinion a core strategy for Anthropic.

Right now, technologists are rushing headlong into this world. That is entirely justifiable. They see an immediate set of largely technical problems to be solved.

However, companies doing even a cursory analysis of their strategy will quickly ask if they really want to be mere conduits for an LLM or organisations that deliver products that leverage the capabilities of LLMs.

My bet is very much on leveraging the capabilities of LLMs. But we have six months to a year of the LLMs as the Platform Core to go through first.