Inspiration
The global loan market now exceeds $2tn across US and European leveraged loans, with a further $2tn+ in private credit. Naturally, stress events are also recurring realities.
This project was inspired by a LinkedIn post from the LMA, where the LMA organised a panel discussing the coming maturity wall and the question loan agents and security agents face.
Combined with our own backgrounds in credit, credit trading, and billion-dollar restructurings, we have seen agents pulled in different directions by lenders and borrowers, often with limited tooling and high personal risk.
What it does
RxOS is an AI-powered operating system built for loan and security agents in stressed and distressed situations.
RxOS was designed from the agent’s seat, focusing on the operational tasks agents are actually responsible for in stressed and distressed situations.
At its core, RxOS allows agents to: (a) Track lender holdings and voting, recognising that positions change frequently due to transfers and sell-downs (also showing lenders' exposure in a particular tranche and their past behaviour and economic incentive); (b) Generate and share security and facility summaries with working parties on a live basis; (c) Send notices and distribute documents directly from the platform, maintaining a clear audit trail; (d) Prepare agent-side outputs, with a roadmap toward auto-drafting standard forms and documents required in an agent capacity.
On top of this, we layered multi-modal simulation, allowing agents to model and compare: (a) What the borrower is pushing for; and (b) How different lenders are likely to behave, based on position, history, and incentives; and (c) What those competing inputs mean for the agent’s duties and risk profile.
From there, RxOS enables agents to take context-aware actions.
Challenges we ran into
In stress situations, ambiguities (especially contractual drafting ambiguities) surface quickly, and agents are expected to act with speed while managing legal and reputational risk.
The hardest challenge was designing a system that could reflect real-world ambiguity and conflicting instructions, without oversimplifying them or forcing false certainty.
We also had to ensure the system feels trustworthy and intuitive to users operating under pressure, where mistakes have serious consequences.
Accomplishments that we're proud of
We built an end-to-end flow that moves from knowledge consolidation to simulation to execution, demonstrating that AI can support judgment and coordination, not just automation.
What we learned
In large, regulated financial markets, the value of AI is not novelty. It is clarity, explainability, and workflow fit. Agents do not need another dashboard. They need systems that help them navigate uncertainty, justify decisions, and act confidently when stakeholders disagree.
What's next for RxOS
Next, we plan to deepen behavioural modelling across lender types, expand coverage into private credit and syndications, and further develop communication and execution tooling.
Our goal is to position RxOS as core infrastructure for agents managing stress and distress across the $2tn+ global loan market,
Log in or sign up for Devpost to join the conversation.