Covenant Sentry

Inspiration

The idea for Covenant Sentry was born from a terrifying realization about modern banking: The 90-Day Blind Spot.

We learned that when a bank approves a loan on Day 1, they typically don't receive updated financial data until the borrower files a quarterly report on Day 90. This creates a 45-day "Silence Gap" where a borrower could lose a major client or drain their cash reserves, and the bank would have absolutely no idea. We realized that Loan Officers are essentially driving a trillion-dollar car while looking only in the rearview mirror.

We asked ourselves: Why are we relying on stagnant PDF reports in an age of real-time APIs? We wanted to build a "Smoke Detector" for loan defaults—a system that alerts the bank the moment a risk occurs, not 3 months later.

What it does

Covenant Sentry is a desktop-native "Risk Command Center" for Loan Officers. It replaces passive quarterly reporting with active, real-time defense.

• Real-Time Monitoring: Instead of static PDFs, the app ingests live simulated accounting feeds to monitor key financial health metrics (Cash Balance & DSCR) every 5 seconds.
• Always-On Background Defense: It runs silently in the System Tray, acting as a 24/7 guardian. Even if the dashboard window is closed, the "Rules Engine" continues to process data in the background.
• Instant Breach Alerts: The moment a borrower crosses a risk threshold, the system fires a native OS notification to the Loan Officer's desktop.
• One-Click Remediation: Clicking an alert opens the "Breach Protocol", an interactive workflow that allows officers to instantly generate default notices, call the CFO, or request waivers.

How we built it

We architected the application using Electron to achieve true desktop integration and security.

• Core Engine: We built the "Rules Engine" and "Mock Data Stream" inside the Electron Main Process. This was a crucial architectural decision because it ensures the monitoring logic runs independently of the UI window.
• Frontend Interface: The visual dashboard is built with React and Vite for high performance, utilizing Recharts for the real-time sparklines and Tailwind CSS to achieve the "Fintech Midnight" aesthetic.
• Secure Bridge: We used Electron's IPC (Inter-Process Communication) to securely bridge the data from the background engine to the React UI, ensuring a strict separation of concerns.

Challenges we ran into

• The "Invisible" App: Getting the application to run correctly in the background when the window was closed was tricky. We had to manage the Electron Tray API and ensure the "Rules Engine" loop didn't get suspended by the operating system's power-saving modes.

• Visualizing Risk: We initially struggled to make the data intuitive. A company might have high cash but a failing DSCR (Debt Service Coverage Ratio). We used LaTeX math modeling to define our critical thresholds:

• Native Notifications: Testing native OS notifications across different operating systems required careful handling of permissions and App IDs to ensure the alerts were delivered reliably.

Accomplishments that we're proud of

• The "Always-On" Architecture: We are most proud of decoupling the UI from the Logic. Being able to close the window and still receive a notification when a borrower crashes felt like a real "Enterprise Grade" achievement.
• The Privacy Sandbox: By processing all data locally on the device (Electron Main Process) rather than sending it to a cloud server, we solved a major data sovereignty hurdle that banks face.
• The Breach Protocol Modal: Turning a passive "Red Alert" into an active workspace where users can fix the problem felt like a huge UX win.

What we learned

• The Cost of Latency: We learned that in finance, "Time is Money" is literal. A reactive cleanup of a defaulted loan is 10x more expensive than a proactive intervention.
• Desktop vs. Web: We gained a deep appreciation for the power of desktop apps. While the web is great, sometimes you need the raw power and system-level access (System Tray, Notifications) that only an Electron app can provide.

What's next for Covenant Sentry

• Modular "Rule Packs": We plan to expand the engine to support different industries. A "Real Estate Pack" could monitor occupancy rates, while a "Healthcare Pack" could monitor patient volume.
• Live API Integration: Currently, we use a robust Mock Data Generator. The next step is to integrate directly with real-world accounting APIs like QuickBooks Online and Xero to ingest live borrower data.
• AI Forecasting: Implementing a simple ML model to predict when a company will breach a covenant based on their current "burn rate," giving Loan Officers even more lead time.

Built With

• electron
• node.js
• react
• recharts
• tailwind
• typescript
• vite