Philly Property Scout.

Looking to buy a multi-family property in Philadelphia on a $40,000 down payment — but the Philly market moves fast, listing quality varies wildly, and manually running cap rate / house-hack math on every duplex would take hours.

Three financing paths needed to be modeled simultaneously for every listing: NACA (0% down, owner-occupancy, 6.5%), FHA (3.5% down, available now, 7.25%), and a conventional investment loan (20% down, 7.5%). A property that looks expensive at list price might save $2,000/month on housing via house-hacking.

The agent needed to surface the signal, not the noise — filtering fixer-uppers, sketchy zips, and impossible numbers before they ever reached the spreadsheet.

The system follows the WAT framework — Workflows define what to do, the Agent decides when and how, Tools execute deterministically. This keeps AI reasoning focused on orchestration while Python handles the math.

Separating probabilistic reasoning from deterministic execution means the scoring model is testable, auditable, and stays accurate no matter how many steps chain together.

Every property is scored 0–100 on five weighted factors, then ranked. The weights are calibrated to a house-hack-first strategy — cash-on-cash return and cap rate are the primary drivers, but house-hack savings matter more than raw price-per-sqft.

First live scan processed 145 raw listings and surfaced 20 investment-grade properties after filtering. The top result — a $199,900 triplex in West Philly — scored a near-perfect 96.8/100 with a 22.26% cap rate and three viable financing paths all within the $40k budget. All three scenarios were feasible simultaneously, which almost never happens at this price point in a major city.

The interactive HTML report generated by export_report.py runs locally in the browser. Sidebar shows all 20 scored properties sorted best → worst. Clicking any property loads the full investment breakdown on the right. Below is the live output from the April 17, 2026 scan.