MCP Protocol Enables Cross-Model Integrations
by Mike Krieger on June 5, 2025
Anthropic's MCP Protocol: Turning Integrations into a Universal AI Standard
Anthropic's Multi-modal Context Protocol (MCP) represents a breakthrough approach to solving one of AI's fundamental challenges: providing models with the right context to deliver useful answers.
Mike Krieger, CPO at Anthropic, frames AI product utility as a three-part equation: model intelligence, context/memory, and applications/UI. While Anthropic's research team focuses on improving model intelligence, MCP specifically addresses the critical middle piece - context and memory.
The genesis of MCP came from a practical problem the team faced. They were building integrations for Claude but found themselves rebuilding each one from scratch in a non-repeatable way. Two engineers, Justin and David, proposed a revolutionary idea: what if instead of building proprietary integrations, they created an open protocol that would allow developers to build integrations once that could be used across multiple AI models?
This approach inverted the traditional dynamic. Rather than asking developers to build integrations specifically for Claude (where Anthropic, as the challenger brand, would struggle to gain adoption), they created a standard that benefits the entire ecosystem. The protocol allows Claude and other AI models to access context from various applications, making integrations built for one model usable by Claude, ChatGPT, Gemini, and others.
The strategy has gained significant traction with key industry players. Toby at Shopify and Kevin Scott at Microsoft have become champions for MCP. Microsoft's recent announcement integrating MCP into their OS demonstrates how quickly this approach is being adopted.
The power of MCP lies in its ability to transform how AI accesses context. Instead of limited information, an AI using MCP can access Slack conversations, Google Drive documents, and other relevant information sources to provide more informed responses. For example, when reviewing product strategy, Claude can now access internal documents, Slack conversations, and Google Drive files to provide much more valuable feedback.
Krieger envisions a future where "everything is an MCP" - where all digital tools and services expose their functionality through this protocol, making them accessible to AI models. This would make everything scriptable, composable, and consistently usable by AI, dramatically expanding what's possible.
The approach represents a classic "commoditize your complements" strategy. By creating an open standard for integrations rather than building them as proprietary features, Anthropic accelerates ecosystem development while focusing on their core strength - building great models. This allows the entire AI industry to move faster while still benefiting Anthropic's position.
For product teams, MCP demonstrates how creating open standards can sometimes be more powerful than building closed systems, especially when you're not the dominant player. By solving a universal problem in an open way, Anthropic has created a growth vector that extends far beyond what they could build themselves.