Patented technology gives AI agents a structured understanding of how code changes ripple across interconnected systems — before any code is written.
London, 1st October 2026 - Â鶹´«Ã½Ó³», a UK-based fintech company delivering delivering mission-critical infrastructure for transfer agency and digital assets, has been granted a United States patent for a system that maintains contextual relationships and operational compatibility in codebases deployed across multiple platforms.
The patent US 12,710,935 covers an AI-driven software development system that understands how a change to code in one platform affects every connected platform. It uses a multi-dimensional knowledge graph and a team of specialised AI agents, an "agent swarm", to map dependencies, learn from past incidents, and generate code that preserves interoperability across systems. Â鶹´«Ã½Ó³» is pursuing other related intellectual property assets.
The system is especially relevant to financial services and transfer agency environments, where organisations manage 15–20 years of transactional data across legacy and modern platforms. A single code change in one environment can silently break processing in another — causing data loss, compounded errors, or system-wide performance degradation that may not surface for weeks.
"Most AI coding tools and agents are very confident strangers. They'll happily change something without knowing what else they might break. What we've patented gives the AI a proper understanding of the system before it starts making changes, how everything connects, what's changed before, and where problems have happened or could happen, in a deterministic way." says Huw Foley, Chief Technology Officer, Â鶹´«Ã½Ó³»
â€
â€The Problem
Current AI code generation tools, including large language models and code assistants, search for code by text similarity. They find a piece of code that matches a description, modify it locally, and deploy it. They don't understand what else that code connects to, what depends on it, or what might break downstream.
Version control systems like Git store code but discard the contextual relationships between code modules. When a change is made, the connections that held the system together are lost. In complex, multi-platform environments, particularly in financial services, this leads to operational breakdowns that are expensive, hard to trace, and sometimes not detected until they've caused real damage.
The Solution
The patented system works differently. Rather than relying on a single AI model to generate code blindly, it deploys a coordinated team of AI agents, each with a specific role. Together, they build and maintain a living map of the entire codebase, a multi-dimensional knowledge graph, that captures every relationship between code modules across every platform.
Why it matters?
For financial services organisations, the stakes are high. Transfer agency data migrations typically require teams of five to six full-time specialists working for 9–12 months, at significant cost. When those projects end, the knowledge leaves with the consultants.
The patented system captures that knowledge permanently. Every decision, every incident, every successful mapping becomes part of the knowledge graph, an institutional memory that compounds over time and makes every future change faster, safer, and more predictable.
More broadly, the patent addresses a fundamental limitation of current AI coding tools. As AI agents increasingly write and modify production code, the ability to understand context, not just generate text, becomes the difference between infrastructure that scales and infrastructure that fragments.
â€
Patent reference: US Patent — claims priority to UK Patent Application 2507202.6, filed 9 May 2025.
Title: System and Method of Maintaining Contextual Relationships and Operational Compatibility in Codebase Functionality Employed Across Multiple Platforms.
â€
