megapredict

The prediction-market screener. One search across every major venue — Polymarket, Kalshi, Limitless, and more — with cross-venue odds comparison and honest price-gap detection.

All market data comes from the PMXT Router (read-only). megapredict builds no ingestion, normalization, or matching of its own — it's a thin Next.js 16 app that renders PMXT's unified schema.

Stack

• Next.js 16 (App Router, Turbopack, Cache Components / use cache)
• TypeScript, React 19
• Tailwind CSS v4
• pmxtjs — the sole data source, accessed server-side only

Getting started

npm install
cp .env.example .env        # then add your PMXT key
npm run dev                 # http://localhost:3000

Get a key at pmxt.dev/dashboard (starts with pmxt_live_) and set it in .env:

PMXT_API_KEY=pmxt_live_xxxxxxxxxxxxxxxxxxxxxxxx

The key is server-side only and is never exposed to the browser. Without it, the app renders a setup hint instead of crashing.

Validate the PMXT integration

PMXT_API_KEY=pmxt_live_xxx npm run validate:pmxt -- "election"

This calls router.fetchMarkets(...) directly and prints unified-schema rows. The home page renders the same fetchMarkets call as a Server Component.

Architecture

app/                         Routes (Server Components + route handlers)
  page.tsx                   Search landing — renders fetchMarkets (RSC)
  screener/page.tsx          Matched-cluster screener table
  markets/[id]/page.tsx      Cross-venue comparison view
  api/                       Client-driven route handlers (search/screener/prices/related)
components/                  Presentational + a few client islands
lib/pmxt/
  client.ts                  Server-only Router (the single PMXT chokepoint)
  queries.ts                 Cached query functions (`use cache` + cacheLife/cacheTag)
  types.ts                   Serializable DTOs + mappers from the unified schema
  policy.ts                  Gap policy (identity + confidence ≥ 0.9 = actionable)
lib/format.ts                Number / probability / freshness formatting

Caching (tuned to balance freshness vs. PMXT rate limits):

Data	TTL	Why
Search results	~30–45s	Catalog updates continuously
Matched clusters	~2–3 min	Cluster membership changes slowly
compareMarketPrices	~10s	Prices drive arb credibility — keep fresh

False-arb protection: a cross-venue gap is only shown as actionable when the relation is identity and confidence ≥ 90%. Everything else is labelled "related — verify rules" with PMXT's reasoning.

Scripts

• npm run dev — dev server
• npm run build — production build
• npm run start — serve the production build
• npm run typecheck — tsc --noEmit
• npm run validate:pmxt — live PMXT integration check