PayPilot AI is a cash-flow credit copilot for online sellers and freelancers. The question behind it is practical: can a person understand their income history and compare financing offers without handing an AI agent unrestricted control over money?
The project reads twelve months of PayPal sandbox income and turns it into a PayScore between 300 and 850. It also highlights the drivers behind the score, surfaces payment gaps, and helps compare the real cost of an offer rather than only its advertised rate.
Keep analysis separate from action
The boundary I wanted to make clear is that an assistant can analyze and explain, but money movement should remain behind explicit borrower approval. The README describes an AI copilot with tools and an approval step before a payout. The demo uses synthetic personas and PayPal sandbox data: it is a prototype, not a lending service, and it does not move real money.
The product is organized around a few connected ideas: PayScore with explainable drivers, an Offer Guard that compares the real cost of an offer, and an Income Passport for sharing income evidence. This makes the demo about more than a single score. It is an exploration of how a financial assistant can make its reasoning visible and put a human confirmation point before consequential actions.
Watch the walkthrough
I recorded a 2-minute-50-second Paypilot demo showing the project flow. You can also explore the live demo or read the source on GitHub.
The main lesson for me is that a useful financial copilot needs more than a confident answer. It needs to explain where its numbers come from, distinguish a sandbox from real financial activity, and make approval explicit before anything can happen.
Top comments (0)