Reference Architecture for AI Agent Access to SFTI APIs

Objective of the project

The project will elaborate a white paper to be shared with the SFTI community containing the following key aspects:

  • Big Picture: Positioning agent-based interactions within the broader Open Finance ecosystem, illustrated by the Open Pension use case
  • Technical Enablement: Defining a reference architecture to ensure secure API access for agents, including among others an MCP facade, enhancing API specification with context for LLMs/agents and secure access/consent handling
  • Recommendations and potential next steps

Why this topic now?

Trend 1: AI-Agents

With the rapid progress of generative AI, intelligent agents are emerging that can independently plan and act. In finance, they unlock new opportunities such as personalized assistants or automated advisory services – provided they have secure access to high-quality data.

Trend 2: Open Finance

Initiatives like Common API from SFTI are laying the technical foundation in Switzerland for standardized and secure data access and new business models.

At the intersection of these two trends lies the Model Context Protocol (MCP) – a new and simple way to bring them together.

As described by its authors:

“MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.” (Source)

Process

The core-project team will work out drafts of the architecture artefacts and the white paper. These drafts are then iterated and validated with the help of a sounding board. The focus will be on the technical, architectural and security aspects of the reference architecture.

The SFTI Reference Architecture for AI Agent Access to Open APIs core-project team consists of representatives from SFTI and Acrea.

The project is co-led by Marco Seiz, Principal Consultant at Acrea and Christof Dornbierer, Partner at Acrea.