Brocrunch

Brocrunch is an agentic DeFi trading platform that lets users deploy autonomous AI-powered trading agents to execute strategies across decentralized exchanges — so you can stop staring at charts and start living your life.

Users configure agents with specific strategies (momentum scalping, mean reversion, cross-DEX arbitrage, yield farming rotation), set risk tolerances, and allocate capital. The agents then trade 24/7 on the user's behalf, executing with precision that human traders simply can't match at 3am when they're doom-scrolling Twitter instead of watching their liquidation price.

The platform features a real-time dashboard with portfolio tracking, PnL analytics, a live 30-day performance chart, individual agent performance sparklines, and a full trade history table. Users can deploy new agents, pause/resume existing ones, and monitor win rates and drawdowns — all from a minimal, monochromatic interface designed to keep the focus on what matters: the numbers.

Built for degens who've evolved past clicking buttons. Not financial advice. Probably.

Brocrunch is built on Next.js 15 with the App Router, using React Server Components and Server Actions for a fast, SEO-friendly experience with minimal client-side JavaScript where possible.

The backend is powered by Supabase — we use Supabase Auth for email/password authentication with session management handled via @supabase/ssr and Next.js middleware that refreshes tokens on every request. The database layer uses four core PostgreSQL tables (profiles, agents, trades, portfolio_snapshots) hosted on Supabase's managed Postgres, with a service-role API endpoint (/api/seed) that populates realistic mock trading data for demo purposes.

The frontend is intentionally minimal — monochromatic black and white with subtle gray accents. No component library; everything is hand-rolled with Tailwind CSS v4. The landing page features a CSS-animated ticker tape, a faux terminal component that "types" real agent commands, and a grid background overlay for that Bloomberg-terminal-meets-hacker aesthetic. Dashboard charts are rendered as inline SVGs with computed polyline paths rather than pulling in a charting library — keeping the bundle lean.

One notably hacky thing: the dashboard runs in a "demo mode" with hardcoded mock data when there's no authenticated session, so judges can explore the full UI without signing up. The mock data generation in the seed endpoint uses realistic crypto pair pricing, randomized PnL distributions, and strategy-specific trade patterns to make the demo feel authentic. The agent "sparklines" on the dashboard are generated from random walks seeded per-agent, giving each one a unique but plausible performance curve.

Authentication flow uses Supabase's PKCE flow with a /auth/callback route that exchanges codes for sessions, and the middleware layer ensures session cookies stay fresh across all routes without requiring client-side token refresh logic.