GreenGeniusAI

An AI-powered automated trading platform — a Claude-driven trading engine analyzes stocks and crypto around the clock, explains every decision in plain English, and can execute trades through a real brokerage (Alpaca), all wrapped in a Next.js dashboard with a $8.99/mo Stripe subscription and an early-stage mobile app.

What it is

Three pieces in one repo:

• backend/ — a FastAPI service. Its core is services/ai_trader.py, an AITradingEngine that prompts Claude with live market data, technical indicators, and portfolio context, and gets back a structured trade decision (action, size, stop-loss, and a plain-English explanation). services/alpaca_service.py executes trades through Alpaca's brokerage API; services/market_data.py pulls live quotes and computes indicators (RSI, MACD, Bollinger Bands, ATR, …) from Polygon.io. APScheduler runs the engine on a real cadence — see the scheduled jobs table below.
• website/ — the Next.js 14 dashboard: onboarding, checkout, a trading dashboard (bots, watchlist, risk, news, derivatives, perpetuals, IPOs, alerts), and a Supabase-backed auth layer. Currently gated behind a private access code rather than open subscription signup (see app/access).
• mobile/ — an Expo / React Native app skeleton (tab navigation scaffolded, screens not yet built out).

This is a working prototype, not an audited production trading system — read LAUNCH_GUIDE.md for the operator's own checklist of what's still needed (legal/compliance, real payment go-live, etc.) before it could take real customer money.

Features

• AI trade analysis — POST /trading/analyze runs a full Claude-driven analysis on any ticker and returns a structured decision with reasoning.
• Autonomous scheduling — a market scan every 20 minutes during market hours, an after-hours deep analysis, and a crypto scan every 30 minutes, all via APScheduler.
• Broker execution — real Alpaca paper/live trading integration (ALPACA_PAPER toggles between them).
• Subscription billing — Stripe checkout, billing portal, and webhook handling.
• Live dashboard — positions, orders, P&L, risk, and a rotating multi-thousand-ticker news scanner, all reading from a live Alpaca account.

Tech stack

Layer	Choice
Backend	FastAPI, APScheduler, alpaca-py, anthropic, Polygon.io via httpx, pandas + ta for indicators
AI	Claude (Anthropic API) — the trading engine's entire decision layer
Frontend	Next.js 14 (App Router), React 18, Tailwind CSS, Zustand, lightweight-charts / Recharts
Auth & DB	Supabase
Payments	Stripe
Mobile	Expo / React Native (early scaffold)

Quickstart

Backend:

cd backend
pip install -r requirements.txt
cp ../.env.example .env   # fill in ANTHROPIC_API_KEY, POLYGON_API_KEY, ALPACA_*, etc.
uvicorn main:app --reload --port 8000

Website:

cd website
npm install
cp ../.env.example .env.local
npm run dev   # http://localhost:3000

Mobile (Expo):

cd mobile
npm install
npm start

See LAUNCH_GUIDE.md for the full account/API-key setup checklist (Anthropic, Polygon, Alpaca, Stripe, Supabase) and .env.example for every variable the app reads.

Architecture

flowchart LR
    subgraph Frontend
        WEB["Next.js dashboard\nwebsite/app"]
        MOB["Expo mobile app\nmobile/app"]
    end

    subgraph Backend["FastAPI backend"]
        AUTH["routers/auth"]
        TRADE["routers/trading"]
        PORT["routers/portfolio"]
        SUB["routers/subscription"]
        WH["routers/webhooks"]
        ENGINE["AITradingEngine\nservices/ai_trader.py"]
        MKT["MarketDataService\nservices/market_data.py"]
        ALP["AlpacaService\nservices/alpaca_service.py"]
        SCHED["APScheduler\nmarket scan / after-hours / crypto scan"]
    end

    subgraph External
        CLAUDE["Claude (Anthropic API)"]
        POLY["Polygon.io"]
        ALPACA["Alpaca Markets"]
        STRIPE["Stripe"]
        SUPA["Supabase"]
    end

    WEB --> AUTH
    WEB --> TRADE
    WEB --> PORT
    WEB --> SUB
    MOB --> TRADE
    TRADE --> ENGINE
    SCHED --> ENGINE
    ENGINE --> CLAUDE
    ENGINE --> MKT --> POLY
    ENGINE --> ALP --> ALPACA
    SUB --> STRIPE
    WH --> STRIPE
    AUTH --> SUPA

Loading

Project structure

backend/
  main.py                  FastAPI app, CORS, APScheduler jobs
  routers/                 auth, portfolio, trading, subscription, webhooks
  services/
    ai_trader.py            Claude-driven trading decisions
    market_data.py           Polygon.io quotes + technical indicators
    alpaca_service.py        broker execution (paper/live)
    stripe_service.py        subscription billing
  models/trade.py           trade decision / action / asset-class types
website/
  app/                      marketing, onboarding, checkout, dashboard/*
  lib/                      alpaca client, auto-trade, signal engine, security, sessions
mobile/
  app/(tabs)/               Expo Router tab scaffold
mcp-server/                 MCP server exposing this repo to AI assistants (see below)
LAUNCH_GUIDE.md             operator checklist: accounts, compliance, go-live steps

MCP Server

mcp-server/ is a working Model Context Protocol stdio server that reads this repository directly — no hardcoded data — so an AI assistant can introspect the real backend, scheduler, and dashboard instead of guessing at them.

Tool	What it does
get_api_routes	Every FastAPI endpoint parsed from backend/routers/*.py
get_scheduler_jobs	Every APScheduler job in backend/main.py, with its real cron/interval schedule
get_dashboard_pages	Every page route under website/app/
get_env_reference	Every variable in .env.example, grouped by section
search_code	Regex search across all tracked source files
read_file	Read any file in the repo by relative path

cd mcp-server
npm install
claude mcp add greengeniusai -- node mcp-server/server.mjs

See mcp-server/README.md for details.

License

MIT — see LICENSE.