A team of specialized AI agents that analyze markets, build trade setups, validate risk, and execute crypto-futures trades in real time — with a live dashboard for portfolio and agent activity.
CryptAI is an autonomous crypto-futures trading system built on a multi-agent architecture. Instead of one monolithic bot, it splits the job across specialized agents — data, analysis, strategy, risk, execution, and memory — orchestrated with LangGraph. Each agent has a single responsibility and hands structured output to the next, so decisions are traceable and every trade passes through a deterministic risk gate before a single order is placed.
Market reasoning combines computational analysis (Smart Money Concepts, ICT, multi-timeframe confluence, classic indicators) with LLM reasoning (Claude + GPT), and every action is streamed to a real-time Next.js dashboard.
⚠️ Research & educational project. Trading crypto derivatives carries substantial risk. CryptAI defaults to testnet / paper trading with execution disabled — enable live trading at your own risk. Nothing here is financial advice.
Live app: https://cryptai-app.vercel.app
Agents run as a pipeline; the risk gate is a hard stop between a proposed setup and execution.
┌────────────┐ ┌──────────────┐ ┌───────────────┐ ┌────────────┐ ┌───────────────┐
│ Data Agent │──▶│ Analysis │──▶│ Strategy │──▶│ Risk Agent │──▶│ Execution │
│ live feeds │ │ SMC/ICT+LLM │ │ trade setups │ │ validate ✋ │ │ place & watch │
└────────────┘ └──────────────┘ └───────────────┘ └────────────┘ └───────────────┘
│ │ │
└─────────────────────── Memory Agent ◀───────────────┴──────────────────┘
trade history · analytics · learning
| Agent | Responsibility |
|---|---|
| Data Agent | Fetches and distributes live market data from Binance Futures (REST + WebSocket streams). |
| Analysis Agent | Orchestrates market analysis — SMC/ICT structure detection, multi-timeframe confluence, and LLM reasoning over the computed context. |
| Strategy Agent | Turns analysis into precise swing/scalp trade setups: entries, stops, targets, confluence score, and risk/reward. |
| Risk Agent | Validates every setup against a rules engine, circuit breaker, correlation and portfolio-heat checks, and drawdown limits — deterministic calc backed by an LLM risk advisor. |
| Execution Agent | Places and monitors orders via exchange APIs, with error handling, retries, and position reconciliation. |
| Memory Agent | Persists trade history and performance analytics, and surfaces learnings from past trades via a vector store. |
- Six specialized agents orchestrated with LangGraph — traceable, single-responsibility decisions.
- Smart Money Concepts & ICT structure detection plus multi-timeframe confluence scoring.
- Hybrid reasoning — deterministic quant signals combined with LLM judgment (Claude + GPT).
- Deterministic risk gate — position sizing, portfolio heat, correlation, drawdown protection, and a global circuit breaker before any order.
- Real-time execution on Binance/BingX with WebSocket order monitoring and position reconciliation.
- Live dashboard — portfolio, open positions, live candles, and per-agent activity in real time.
- Paper-trading mode — full simulation loop with no capital at risk (default).
- Persistent memory & analytics — trade journaling, performance metrics, and vector-backed retrieval.
| Layer | Technologies |
|---|---|
| Agents / AI | LangGraph · LangChain · Claude (Anthropic) · GPT-4o (OpenAI) |
| Analysis | pandas · NumPy · SciPy · TA-Lib · pandas-ta · custom SMC/ICT detectors |
| Market data / exec | ccxt · python-binance · WebSockets · BingX / CoinDCX |
| Backend | Python 3.11 · FastAPI · asyncio · Pydantic · Loguru · Prometheus |
| Data stores | PostgreSQL (SQLAlchemy + Alembic) · Redis (streams/state) · Pinecone / ChromaDB (memory) |
| Frontend | Next.js · React · TypeScript · Tailwind CSS · Radix UI · TanStack Query · Zustand · lightweight-charts · Supabase (auth) |
| Infra | Docker Compose · Vercel (frontend) · Render (backend) |
CryptAI/
├── crypto-trading-agent/ # Python backend — the agent system
│ ├── src/
│ │ ├── agents/ # data · analysis · strategy · risk · memory
│ │ ├── analysis/ # SMC/ICT detectors, indicators, MTF analyzer
│ │ ├── strategy/ # setup builder, confluence, sizing, R:R
│ │ ├── risk/ # rules engine, circuit breaker, portfolio heat
│ │ ├── execution/ # order placement & monitoring
│ │ ├── memory/ # trade history & vector retrieval
│ │ ├── core/ · data/ · api/ · utils/
│ ├── alembic/ # DB migrations
│ ├── docker-compose.yml # Redis + PostgreSQL
│ └── requirements.txt
├── frontend/ # Next.js dashboard
└── docs (SETUP.md, ARCHITECTURE_DEEP_DIVE.md, PAPER_TRADING_GUIDE.md, …)
- Python 3.11+, Node.js 18+, Docker Desktop
- API keys: Anthropic and OpenAI (required); Binance (market data); BingX (optional, live trading); Pinecone (optional, memory)
git clone https://github.com/sahit1011/CryptAI.git
cd CryptAI/crypto-trading-agent
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env # then fill in your keys
docker-compose up -d # starts Redis + PostgreSQL
python init_database.py # run migrations / seed schemaKeep the defaults USE_TESTNET=true and ENABLE_EXECUTION=false for safe paper trading. See crypto-trading-agent/PAPER_TRADING_GUIDE.md.
cd frontend
npm install
# create .env.local with your Supabase keys + backend API URL, e.g.
# NEXT_PUBLIC_SUPABASE_URL=...
# NEXT_PUBLIC_SUPABASE_ANON_KEY=...
# NEXT_PUBLIC_API_URL=http://localhost:8000
npm run dev # http://localhost:3000All backend settings live in crypto-trading-agent/.env (see .env.example):
| Key | Purpose |
|---|---|
ANTHROPIC_API_KEY / OPENAI_API_KEY |
LLM reasoning |
BINANCE_API_KEY / BINANCE_SECRET_KEY |
Market data |
BINGX_API_KEY / BINGX_SECRET_KEY |
Live execution (optional) |
REDIS_URL / POSTGRES_URL |
State & persistence |
USE_TESTNET / ENABLE_EXECUTION |
Safety switches (true / false by default) |
🔐 Never commit real keys.
.envis gitignored — only.env.example(placeholders) belongs in git.
Detailed guides: SETUP.md · crypto-trading-agent/ARCHITECTURE_DEEP_DIVE.md · BACKEND_STARTUP_GUIDE.md
CryptAI is provided for research and educational purposes. Automated trading of leveraged crypto derivatives can lead to significant financial loss. Use testnet/paper mode unless you fully understand the risks, and never trade capital you can't afford to lose. The author accepts no liability for financial losses. This is not financial advice.
Released under the MIT License.