PawMatch

PawMatch is a smart layer on top of PetPoint (the shelter system of record) plus a Bumble-style adoption app. Shelters' PetPoint data is messy, batch-exported monthly, and unusable for adopters - so pets get adopted via Instagram instead. PawMatch pulls PetPoint records through a normalization layer that maps raw export columns to a clean schema, flags quality issues per field (missing photos, blank spay/neuter, unmapped intake subtypes, stale records), and computes a trust score that gates publishing - making shelter data trustworthy and accessible in near real time.

Adopters onboard in a 4-step wizard, then swipe through compatibility-ranked pets (weighted score: city, distance, home fit, energy, experience). A match files an application to the shelter automatically. Every pet has an AI persona chat that answers from its real records (medical, quirks, backstory) while naturally gathering the adopter's info for the shelter - and then the Arcade.dev agent finishes the job: it books a real meetup on Google Calendar (real OAuth via Arcade's actions runtime) and emails confirmation via Gmail. Shelters get a desktop 'Shelter OS': live KPIs, AI profile completion that turns messy intake notes into swipeable profiles, AI-prioritized application inbox, and a PetPoint sync console with per-record trust scoring.

Built for the Arcade.dev track: an agent that actually does the work - 4 agents (profile completer, pet persona, application prioritizer, meetup scheduler), all with demo-mode fallbacks so nothing hard-fails. Next.js 16, deployed on Vercel. No PetPoint account needed: the connector ships against a faithful mock of their export API - swap the base URL when credentials arrive.