PayGuard – Cash-Flow Agent

Global AI Hackathon — Powered by Mel | Track 2: Agentic Finance

📝 Existing Project / Open-Source Disclosure

Original Project (Pre-Hackathon): Our starting point was "PayGuard," a static HTML/CSS landing page template we previously built. Before this hackathon, it only contained a basic front-end layout and static dummy tools (like a static fee calculator). It had zero agentic capabilities, no cash-flow logic, and no backend.

What We Built During the Hackathon: During this hackathon, we built the entire Agentic Finance workflow completely from scratch using the Mel terminal. Specifically, our team:

1. Created the cashflow-agent.html dashboard layout and UI.
2. Wrote the js/cashflow-agent.js analytical engine that powers the agent (reads the mock data, calculates burn rate, isolates one-time purchases, and flags anomalies).
3. Built the "Agentic Core" terminal console that outputs the agent's thought process step-by-step.
4. Rewrote the main index.html to integrate the new Cash-Flow Agent.

Everything related to the "Cash-Flow Agent" (the entire Track 2 submission logic) was built during the hackathon build window natively using Mel.

---

🚀 About the Cash-Flow Agent

Startups and freelancers struggle to interpret their monthly cash flow. They see raw transaction data but fail to spot creeping expenses (like escalating marketing or subscription costs) until it's too late.

The Cash-Flow Agent acts as a tireless financial analyst. It follows a strict agentic loop: Read Data → Detect → Investigate → Calculate → Recommend → Flag

Our agent doesn't just display data; it actively compares month-over-month spending, isolates one-time purchases (like a laptop) from recurring burn, and flags specific actions (like "Audit software subscriptions") based on its investigation.

🛠️ Built With

• Mel (AI Harness)
• HTML5, Vanilla CSS, Vanilla JavaScript