quantclaw
v0.1.5
Published
Open-source quant trading superagent harness
Maintainers
Readme
QuantClaw
Open-Source Quant Trading Superagent Harness
From beginner to autonomous quant trader
Documentation | Contributing | Discord
QuantClaw is an open-source quant trading harness that orchestrates 13 AI agents to handle the full trading lifecycle: data ingestion, signal generation, backtesting, live execution, risk monitoring, and reporting.
Two commands. Five minutes. You're trading.
Install
npm install -g quantclaw@latest
quantclaw startOpen http://localhost:3000 in your browser.
If
quantclaw startis not recognized, usenpx quantclaw startinstead. This works without PATH configuration.
Prerequisites
- Node.js 20+
- Python 3.12+
- Ollama (optional, for local AI models)
What Happens
npm install -g quantclaw@latestinstalls everything: Python backend, Node.js sidecar, Next.js dashboardquantclaw startboots all three services in the background- Browser onboarding walks you through language, model provider, search, data sources, broker, and watchlist setup
- You're live with a chat-first trading interface powered by AI agents
quantclaw start # Start all services
quantclaw stop # Stop all services
quantclaw status # Check what's runningModel Providers
QuantClaw supports multiple AI providers. Pick one during onboarding or configure later in Settings.
| Provider | Auth | Models | |----------|------|--------| | Ollama | Local, no key | qwen3, llama, mistral, etc. | | OpenAI | OAuth (subscription) or API key | gpt-5.4, gpt-5.3-codex | | Anthropic | OAuth (subscription) or API key | claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5 | | Google Gemini | Free API key from aistudio.google.com | gemini-3.1-pro, gemini-2.5-flash | | DeepSeek | API key | deepseek-chat, deepseek-reasoner | | xAI | API key | grok-4.20, grok-4-1-fast | | Mistral | API key | mistral-large, mistral-small, codestral | | Groq | API key | llama-3.3-70b, gpt-oss-120b, qwen3-32b | | OpenRouter | API key | 200+ models from all providers | | Together AI | API key | Llama, DeepSeek, Qwen open models |
China users: QuantClaw auto-detects your region and shows Chinese providers (DeepSeek, Qwen, Doubao, GLM, Kimi, ERNIE, Spark, etc.) with Chinese-language UI.
13 AI Agents
Each agent handles a specific part of the trading lifecycle. Assign different models to different agents based on their compute needs.
| Agent | Role | Tier | |-------|------|------| | Scheduler | Coordinates workflows between agents | Light | | Ingestor | Pulls market data and processes feeds | Medium | | Miner | Factor mining with LLM-powered discovery | Heavy | | Backtester | Runs strategy backtests with realistic costs | Heavy | | Researcher | Analyzes markets, papers, and factor performance | Heavy | | Risk Monitor | Monitors drawdowns, exposure, and VaR | Heavy | | Executor | Handles order execution and broker integration | Medium | | Reporter | Generates portfolio summaries and P&L reports | Light | | Trainer | Trains ML models for signal generation | Heavy | | Compliance | Checks trading against regulatory rules | Medium | | Cost Tracker | Monitors API usage and compute costs | Light | | Debugger | Diagnoses system issues and traces errors | Medium | | Sentinel | Always-on market monitoring daemon | Medium |
Market Data
44+ free data sources enabled by default, covering:
- Stocks & ETFs -- Yahoo Finance, Stooq, Finnhub, Twelve Data, Alpha Vantage, and more
- Crypto -- CoinGecko, Binance, CCXT (100+ exchanges), CoinPaprika, Kraken, Coinbase
- Forex -- ExchangeRate-API, Open Exchange Rates, Fixer.io
- Commodities -- Oil Price API, Alpha Vantage, Commodities-API, Metals-API
- Economic Data -- FRED, World Bank, US Treasury, IMF, BLS
- Alternative -- SEC EDGAR, ApeWisdom (Reddit sentiment), FINRA dark pool data
- Kroness -- Built-in unified feed across all markets
China users get AKShare, BaoStock, Tushare, EastMoney, Sina Finance, and other domestic sources.
Add API keys for premium providers anytime -- QuantClaw auto-upgrades routing based on rate limits.
Web Dashboard
Dark-mode, multilingual (English / Chinese / Japanese) dashboard with:
- Chat-first interface -- Talk to your agents, route messages automatically
- Model selector -- Switch providers and models per conversation
- Agent configuration -- Assign models to each agent, auto-assign by tier
- Portfolio overview -- Positions, P&L, equity curve
- Strategy browser -- 55 templates across beginner to advanced
- Backtest runner -- One-click strategy testing
- Risk monitoring -- Drawdown, exposure, VaR limits
- Settings -- Provider OAuth, API keys, language, notifications
Strategy Templates
Write strategies as simple Python files:
class Strategy:
name = "Simple Momentum"
universe = ["AAPL", "MSFT", "GOOGL", "AMZN", "NVDA"]
frequency = "weekly"
def signals(self, data):
scores = {}
for symbol in self.universe:
df = data.history(symbol, bars=25)
if len(df) >= 20:
scores[symbol] = df["close"].iloc[-1] / df["close"].iloc[-20] - 1
return scores
def allocate(self, scores, portfolio):
ranked = sorted(scores, key=scores.get, reverse=True)[:3]
return {s: 1/3 for s in ranked}55 templates included: momentum, mean reversion, pairs trading, risk parity, sector rotation, ML signal, options wheel, and more.
Progression System
QuantClaw grows with you:
| Level | Title | What Unlocks | |-------|-------|-------------| | 0 | Observer | Browse templates, explore backtests | | 1 | Paper Trader | Run backtests, paper trade | | 2 | Strategy Tinkerer | Edit parameters, compare strategies | | 3 | Live Trader | Real broker, live execution, alerts | | 4 | Strategist | Write custom strategies | | 5 | Quant | Factor mining, ML models, deep research | | 6 | Architect | Multi-strategy ensemble, custom plugins |
Architecture
quantclaw start
|
+-- Python Backend (FastAPI, port 8000)
| Agent orchestration, data APIs, OAuth
|
+-- Node.js Sidecar (Express, port 8001)
| OpenAI/Anthropic OAuth proxy (subscription-based)
|
+-- Next.js Dashboard (port 3000)
Chat UI, onboarding, settings, i18nPlugin System
quantclaw install broker-alpaca # Add Alpaca broker
quantclaw install data-polygon # Add Polygon data
quantclaw install engine-vectorbt # Add vectorbt backtest engine
quantclaw install asset-crypto # Add crypto tradingSearch Providers
Agents use web search for market research and news analysis:
Brave Search, Tavily, Exa, Perplexity, Firecrawl, Gemini Search, Grok Search, Kimi, DuckDuckGo (free fallback), SearXNG (self-hosted).
Contributing
See CONTRIBUTING.md for guidelines.
License
MIT License. See LICENSE for details.
