Dedicated money agent
Onboarding provisions a personalized assistant with tone, guardrails, settings, and a compiled system prompt.
P2PALConversational financial layerAI-powered P2P payments demo
P2PAL
A conversational financial layer that turns multi-step P2P transactions into a guided chat stream with an AI agent, editable confirmation cards, and local session history.
Demo stateFake rails
Core flowSend + confirm
AI pathGemini ready
What it proves
A demo that behaves like a production fintech assistant.
P2PAL already demonstrates the end-to-end customer experience: onboarding, agent personalization, intent extraction, editable cards, confirmation gating, receipts, sessions, mobile layouts, and Worker-backed AI plumbing.
Send, request, split
Natural language turns into structured local PaymentIntent cards with required fields and limits.
Confirmation gated
The demo never claims real payment movement. Every action stays fake until confirmed by the user.
Session history
New threads preserve prior transcripts, receipt cards, and active intent state in local storage.
Demo route
Let users test the app without touching real funds.
The demo route opens the same mobile-native app surface. Users can onboard, create an agent, draft P2P actions, edit card details, confirm mock receipts, and switch sessions.
Open /demo
Continue through demo auth
Create your agent
Try: Send $25 to Maya for lunch
Screenshot showcase
Overview of the working demo
These are saved from the current build in `public/screenshots`.
Gemini-ready flow
AI extraction plus confirmation
The chat calls the Worker extraction path first, then falls back locally when keys are not configured.
Chat artifacts
Inline payment history
Send, request, and split cards stay embedded in the thread and restore with local sessions.
Investor demo slice
Editable confirmation cards
Amount, source, recipient, memo, and split participants can be edited before confirmation.
Operational view
Command Center
Wallet, activity, contacts, sessions, and settings sit beside the GPT-style chat surface.
Mobile chat composer
Mobile split receipt
Mobile contact chips
Build stack
Frontend-first, backend-ready.
Next.js, React, Tailwind CSS, and Zustand power the mobile-native frontend.
Cloudflare Worker routes handle onboarding, agent creation, chat, and structured intent extraction.
Gemini 3.1 Flash Lite is wired behind a typed extraction endpoint with local fallback.
Supabase schema is scaffolded for profiles, KYC, connected accounts, agents, sessions, and events.