The Real Story Isn’t the Models—It’s the Infrastructure

While headlines chase the latest Claude or Gemini variant, the actual revolution is happening in the plumbing. The Model Context Protocol (MCP) crossed 97 million installs in March 2026, a milestone that fundamentally reframes what matters in the AI industry right now.

This isn’t hyperbole. In a landscape where major model releases have slowed to a crawl (April 2026 saw no significant open-source releases), infrastructure adoption tells you where the real money and momentum are building.

What’s Actually Happening

The Agentic AI Foundation, formed under the Linux Foundation in December 2025 with Anthropic’s MCP as a founding contribution, represents a deliberate industry pivot. Rather than competing on raw model capability—a game increasingly dominated by well-funded labs—the focus has shifted to making AI systems composable, reliable, and enterprise-ready.

MCP’s 97 million installs mean developers can now:

  • Build standardised connectors between AI models and external tools
  • Create reproducible agentic workflows without reinventing integrations
  • Deploy agents across heterogeneous toolsets at scale

This is less “ChatGPT killer” and more “HTTP for AI agents.”

Why This Matters for Irish and European Builders

Ireland’s strong AI production readiness (41% of organisations moving beyond experimentation, according to recent analysis) positions local builders uniquely to capitalise on this infrastructure moment. Rather than racing to build the next frontier model—a game Irish teams were never going to win against OpenAI or DeepMind—the real opportunity is in the middleware layer.

European AI companies building on standardised agentic infrastructure face lower switching costs, better interoperability with EU regulatory frameworks, and reduced vendor lock-in—exactly what enterprises will demand as the August 2026 EU AI Act transparency deadline approaches.

Practical Implications

For Irish developers and teams:

1. Tooling Over Models: Investment should shift toward building domain-specific agents (healthcare compliance, financial services, supply chain) rather than chasing general-purpose capability.

2. Regulatory Alignment: MCP’s standardisation makes it easier to audit, log, and control AI system behaviour—critical for compliance with EU sectoral regulators managing the August 2026 deadline.

3. Talent Acquisition: Agentic AI infrastructure expertise will command premiums over “generic ML” skills. Teams building MCP integrations will be in higher demand than those fine-tuning models.

Open Questions

The consolidation around MCP raises some thorny questions:

  • Will standardisation stifle innovation? If everyone builds on the same infrastructure, how do differentiation and competitive advantage emerge?
  • Is this sustainable? 97 million installs sounds impressive—but how many represent active, paying deployments versus developers kicking the tyres?
  • What about closed-source models? OpenAI and Google haven’t committed to MCP with the same vigour as Anthropic. Could we see a fork?

The Bottom Line

April 2026 isn’t notable for what was released. It’s notable for what wasn’t released—and what that tells us about where the industry is heading. The age of chasing model leaderboards is giving way to the age of reliable, auditable, composable agentic systems.

For Irish teams, that’s good news. We’re competitive in infrastructure. We’ve built products on standards before. And we understand regulation.

The question isn’t whether you should be thinking about agentic AI. The question is whether you’re thinking about it at the infrastructure layer or still playing catch-up in the model layer.

Source: Foxxe Labs Analysis