What Is MCP and Why Every Financial Institution Should Be Paying Attention

Guest Editorial by Matthew Terry, Chief Technical Officer at Nymbus

If you have not yet heard the term Model Context Protocol, you will soon. And if you work in financial services, it is worth understanding now, because MCP may represent the most practical breakthrough in AI adoption our industry has seen in years.

Matthew Terry, Chief Technical Officer at Nymbus.

For all the excitement around AI in financial services, most institutions have struggled to move beyond experimentation. Yet, nearly half of financial services CEOs identified digital and AI investments as key to success in 2026, with more than 50% saying generative AI will be a primary driver of transformation, according to Gartner's CEO Outlook Survey.

While adoption is widespread, progress has been uneven. McKinsey research shows that the majority of organizations are still experimenting with or piloting AI, with only about one-third successfully scaling it across the enterprise. In banking, that gap between interest and execution is even more pronounced.

AI tools can answer questions, summarize documents, and generate content, but connecting them to the actual systems where banking work happens has been an entirely different challenge. Integrating an AI assistant with a core banking platform, a loan origination system, or a member service portal typically requires custom development work, bespoke APIs, and ongoing maintenance that strains already-thin technology teams. In fact, more than 70% of financial institutions cite fragmented data and infrastructure as key barriers to scaling AI, reinforcing that the challenge is less about AI capability and more about integration.

The result is that even where AI is deployed, its impact remains limited. S&P reports just 9% of financial institutions are using AI in customer- or member-facing applications, highlighting how difficult it is to embed these tools into core workflows. While many institutions report experimenting with AI, far fewer are seeing meaningful, measurable returns.

MCP gives financial institutions the connective layer they need to turn AI experimentation into meaningful operational impact.

What MCP Actually Is

Model Context Protocol is an open standard that creates a common language between AI systems and the tools, data sources, and platforms they need to interact with. Think of it like USB for AI. Before USB, connecting a device to a computer required knowing exactly what kind of port it used, what drivers it needed, and whether it was compatible with your operating system. USB replaced all of that with a single, universal connection. MCP is attempting to do the same thing for AI.

Without a standard like MCP, every AI deployment in banking requires point-to-point integration. You want your AI assistant to look up a member account? Build an integration. You want it to initiate a card freeze? Build another one. You want it to pull transaction history and flag anomalies? Build yet another. Every connection is custom, every update creates risk, and the accumulated complexity becomes a ceiling on what is actually possible.

MCP replaces that architecture with a single, standardized interface. An AI assistant that supports MCP can connect to any MCP server, call approved functions through a consistent protocol, and return results to the user, all without requiring institutions to rebuild the connection every time something changes.

Why This Matters for Credit Unions

For credit unions, the implications are especially important. Financial institutions operate in a heavily regulated environment where control, accountability, and auditability are not optional. That reality has made AI adoption slower in banking than in other industries, and understandably so. Moving faster than your governance structure allows is not a competitive advantage in financial services. It is a liability.

Regulators have made it clear that institutions must be able to explain and govern AI-driven decisions, reinforcing the importance of transparency, auditability, and control. At the same time, concerns around trust, interpretability, and risk management continue to limit broader deployment.

What MCP offers is not just a simpler way to connect AI to core systems; it offers a structured way to do so within institution-defined guardrails. When an MCP server is built with banking compliance in mind, institutions can define exactly which tools are available, which roles can access them, where human approval is required before an action is taken, and how every interaction is logged. The AI does not decide what it can do. The institution does.

That distinction matters enormously. Regulators are not asking financial institutions to slow down AI adoption. They are asking institutions to be able to explain and document it. An MCP-based architecture, built with role-based access controls, full audit logging, and PII protections, gives institutions something to point to when those questions come.

The Day-to-Day Opportunity

The near-term value of MCP in financial services is not about replacing employees or automating decisions. It is about reducing the friction that surrounds everyday work.

A member service representative who currently toggles between four systems to resolve a single inquiry could work through a single conversational interface instead. A fraud analyst who spends an hour gathering account history before beginning an investigation could compress that to minutes.

The opportunity is not trivial. McKinsey estimates that AI could reduce banking costs by 15% to 20% as it becomes embedded into core workflows.

These are not transformational changes in isolation. But compounded across thousands of daily interactions, they represent meaningful efficiency gains, better member experiences, and more time for the work that actually requires human judgment.

Still Early, But Moving Fast

MCP is still an emerging standard. The ecosystem around it is growing rapidly, but financial services adoption is only beginning. McKinsey suggests the greater competitive risk may not be experimenting with AI too aggressively but failing to scale and operationalize AI initiatives while competitors move ahead.

Credit unions that lean into MCP now, even in limited, controlled pilots, will be better positioned to understand its constraints, evaluate vendors building on it, and make informed decisions about where and how to expand its use. Those pilots can begin with read-only use cases, such as retrieving account information, summarizing member interactions, or helping employees navigate policies before expanding into workflows that require approvals or transactional actions. Those that do will also have a stronger opportunity to retain, deepen, and grow member relationships as expectations for AI-friendly banking platforms evolve.

For credit unions in particular, MCP could level a playing field that has historically tilted toward larger institutions with more resources for custom AI development. If the standard matures as predicted, the cost and complexity of connecting AI to core banking functions should fall substantially, making sophisticated AI-assisted workflows accessible to institutions of all sizes.

MCP is worth understanding and planning for now, before larger institutions turn AI-enabled banking workflows into a competitive advantage.

---

Matthew Terry is Chief Technology Officer at Nymbus, a modern core banking platform for U.S. banks and credit unions. Nymbus delivers a cloud-based, highly extensible, full-stack banking platform, empowering community banks and credit unions to accelerate their growth and market positioning.