Key Takeaways

  • MCP can enable AI agents to securely interpret and act on PSD2-compliant banking APIs.
  • It is designed to abstract technical API complexity into model-readable instructions for generative AI.
  • By layering MCP over open banking API standards such as PSD2, banks and fintechs stand to gain faster integration, better governance, and richer user experiences.
  • MCP can turn regulatory compliance into a strategic advantage for intelligent orchestration.

Since the introduction of PSD2 in Europe, open banking has become an essential regulatory framework. Banks and fintechs connect via standardized APIs to initiate payments and access account data, with precise authentication and security rules. This foundation is now essential in Europe for PSD2, and in other formats around the world.

But a central question remains: how can these regulated flows be brought into effective dialogue with new digital uses, notably generative AI, which powers conversational assistants?

This is where Model Context Protocol (MCP) opens up new perspectives.

MCP: a common language for banking APIs and AI

Open banking APIs are designed for human developers: they define data schemas, endpoints, and authorization flows. MCP, on the other hand, aims to enable AI models (LLMs) to interact directly with these systems, by standardizing the way in which a resource or capability is exposed to a model.

Layering MCP “on top of” PSD2 or equivalent standards can create a universal abstraction layer where AI can discover, understand, and manipulate banking services in a secure and contextualized way, without having to re-implement for each regionally-mandated API or technical variation.

Concrete benefits:

Accelerated interoperability

  • An AI embedded in an application can query different European banking providers without worrying about differences in PSD2 implementation.
  • MCP translates the “raw” language of APIs into model-intelligible instructions.

Enhanced security

  • MCP does not override regulatory constraints: it relies on existing authorizations (OAuth2, SCA, etc.).
  • But it does offer better governance of the contexts transmitted to models: only relevant data is shared, with explicit control.

Fluid user experience

  • End-users don’t have to handle complex tokens or consents: they dialogue with an assistant who, thanks to MCP, manages these steps in the background, while complying with regulations.
  • For example, you can ask an AI agent “What is my savings capacity this month?” It would then call the PSD2 APIs via MCP and give a clear, synthesized answer for all accounts in the various establishments.

Innovation through contextualization

  • Beyond simple balance or payment initiation calls, MCP enables generative AIs to mix several contexts: banking data, calendar, budget objectives, savings recommendations.
  • We move from “technical” access to a much richer ecosystem of uses.

Strategic leverage for banks and fintechs

Open banking has already forced traditional players to open up their data. Opening up to MCP is the next step:

  • For banks, it’s an opportunity to offer intelligent conversational interfaces, based on their existing services.
  • For fintechs, it’s an integration gas pedal: instead of managing the complexity of each API, they dialogue with a single protocol, already understood by the models.
  • For users, it’s the promise of an enhanced, simpler, and more proactive banking relationship.

Towards intelligent orchestration of banking services

The PSD2 and open banking standards have laid the foundations for a secure, interoperable ecosystem. But their language remains that of developers to machines… not that of artificial intelligence. MCP provides the missing layer: a contextual protocol that bridges the gap between regulated APIs and conversational uses.

In other words, opening up MCP to open banking means transforming a regulatory obligation into a lever for innovation – and enabling banks to move from APIs to intelligent agents.

Beyond simple access to your MCP server, facilitated by a gateway, it is essential to be able to orchestrate these complex flows with fluid execution across various systems. Axway’s Amplify platform, with its federated API management and integration capabilities, as well as a unified orchestration layer, offers a clear strategic advantage over API-only approaches – especially as your organization relies more and more on AI.

With Amplify Fusion, you can not only create MCP interfaces, but also integrate and orchestrate flows with any backend, in a composable, low-code/no-code way.

Finally, any MCP servers you create can be automatically attached to the Amplify Engage marketplace. Third-party MCP interfaces can be added in just a few clicks, giving you a centralized registry bringing together your AI assets and the more traditional assets of a digital services marketplace.

There are still challenges related to security and agentic AI, as we highlighted in a previous article, but we at Axway will continue to monitor these developments closely to best support you in your transformation.

In a forthcoming post, we’ll look at some case studies combining LLM and financial data connected via MCP.

Architecting trust: how to build secure, scalable, and reliable AI-driven solutions

Share this article