Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for JobProof Pay Agent
The problem
Home-service and gig platforms still release payment on trust, chat screenshots, or slow human review. Fake "done" claims waste money; delayed payouts frustrate good contractors.
The solution
JobProof Pay Agent puts an AI completeness gate in front of PayPal money movement:
- Create a job with amount + contractor PayPal email
- Upload before/after proof photos (or use demo samples)
- Local AI scorer rates completeness (works with zero OpenAI keys for judges)
- On pass, a PayPal agent (
@paypal/agent-toolkit) creates an Orders v2 order for the customer - After sandbox approval, we capture the order and optionally Payout the contractor
What we built
- Next.js App Router MVP with end-to-end sandbox payment flow
- Deterministic local AI + optional OpenAI augmentation
- PayPal Agent Toolkit tools:
create_order,get_order,pay_order - PayPal Payouts API for contractor settlement
- Jobs table for an operator / sponsor-friendly view
- Demo mode when credentials are missing so the UX still demos
How PayPal + AI are used
- AI: local heuristic completeness model over image metadata + job context; optional LLM judge
- PayPal: Orders v2 (customer pay) + Payouts (contractor) orchestrated by an agent toolkit layer
Demo instructions for judges
`git clone https://github.com/icohangar-ops/jobproof-pay-agent cd jobproof-pay-agent cp .env.example .env.local
Add PAYPAL_CLIENT_ID + PAYPAL_CLIENT_SECRET from developer.paypal.com (Sandbox)
npm install npm run dev ` Walkthrough: Home → Create job → Use kitchen samples → Run AI score → Create PayPal order → Approve with sandbox buyer → Success (capture + payout).
Without credentials, the same path runs in demo mode (simulated order/capture IDs) so you can still evaluate UX and AI gating.
Built With
- ai-(local-scorer-+-optional-openai)
- next.js
- paypal-agent-toolkit
- paypal-orders-v2
- paypal-payouts
- tailwind-css
- typescript

Log in or sign up for Devpost to join the conversation.