🚀 AgentQuant

AI-Powered Quantitative Trading Platform

Transform natural language into sophisticated trading strategies with real-time backtesting and interactive visualization

🎯 Overview

AgentQuant is a cutting-edge AI-assisted quantitative trading platform that revolutionizes how traders and developers approach strategy development. By combining advanced language models with sophisticated tooling infrastructure, it transforms natural language trading ideas into executable, backtestable strategies with comprehensive performance analytics.

AgentQuant's intuitive interface - "The Flight Simulator for Trading Strategies"

See AgentQuant in action - from natural language to trading results in seconds!

🚀 Ready to try AgentQuant?Join our waitlist for early access to the platform!

✨ Key Features

🤖 AI-Powered Strategy Generation: Convert natural language into sophisticated trading logic
📊 Real-Time Market Data: Live data feeds with intelligent caching and preprocessing
🔬 Advanced Backtesting: Isolated execution environments with comprehensive metrics
📈 Interactive Visualization: Beautiful charts, KPIs, and trade analysis
🔄 Streaming Progress: Real-time feedback during strategy development
🛠️ Extensible Architecture: Modular design with pluggable components
🚀 Production Ready: Docker-based deployment with Kubernetes support

🏗️ Architecture

System Components

graph TB
subgraph "Frontend Layer"
UI[Web App<br/>Next.js + React]
UI --> |Chat Interface| API[API Routes]
end
subgraph "AI Layer"
AGENT[AI Agent<br/>LangChain + NAT + Qwen]
AGENT --> |Tool Calls| MCP[MCP Server]
end
subgraph "Data & Execution Layer"
MCP --> |Market Data| REDIS[(Redis Cache)]
MCP --> |Code Execution| K8S[Kubernetes<br/>Execution Sandbox]
K8S --> |Results| REDIS
end
subgraph "External Services"
MARKET[Market Data<br/>Yahoo Finance]
LLM[LLM Services<br/>OpenAI/Qwen]
end
API --> AGENT
AGENT --> LLM
MCP --> MARKET
classDef frontend fill:#667eea,stroke:#764ba2,stroke-width:3px,color:#fff
classDef ai fill:#f093fb,stroke:#f5576c,stroke-width:3px,color:#fff
classDef data fill:#4facfe,stroke:#00f2fe,stroke-width:3px,color:#fff
classDef external fill:#43e97b,stroke:#38f9d7,stroke-width:3px,color:#fff
class UI,API frontend
class AGENT,MCP ai
class REDIS,K8S data
class MARKET,LLM external

Technology Stack

Component	Technology	Purpose
Frontend	Next.js 15, React 19, TypeScript	Modern web interface with streaming UI
AI Agent	LangChain, NVIDIA NAT, Python 3.12	Natural language processing and tool orchestration
Tooling Service	FastAPI, MCP Protocol	Unified API for market data and execution
Execution	Kubernetes, Docker	Isolated backtesting environments
Data Store	Redis 7	High-performance caching and state management
Charts	Lightweight Charts, Recharts	Professional financial visualizations
Styling	Tailwind CSS, Radix UI	Modern, accessible design system

End-to-End Sequence (Simplified)

sequenceDiagram
participant User
participant Web as Web
participant Agent as Agent
participant Tools as Tooling Service
participant Store as Data Store
participant Runner as Execution Sandbox
User->>Web: Submit strategy request
Web->>Agent: Send request (streaming)
Agent->>Tools: Register task / preview parameters
Agent->>Tools: Fetch market data (cached)
Agent->>Tools: Generate and run backtest
Tools->>Runner: Execute strategy in isolation
Runner->>Store: Read/write data
Tools-->>Agent: Return metrics and outputs
Agent-->>Web: Summarize results and visuals
Note over User,Web: Frontend Layer
Note over Agent,Tools: AI & Service Layer
Note over Store,Runner: Data & Execution Layer

🚀 Quick Start

📋 Prerequisites

Requirement	Version	Notes
Node.js	18+	With `pnpm` package manager
Python	3.12+	With `uv` package manager
Docker	Latest	& `Docker Compose`
Kubernetes	Latest	For code execution sandbox
Redis	7+	(Included in Docker setup)
LLM API Keys	-	For both generic and coder models

Option 1: Docker Compose (Recommended)

# Clone the repository
git clone https://github.com/nlpquant/AgentQuant.git
cd AgentQuant
# Set up environment variablescd deploy
export LLM_GENERIC_MODEL_NAME='qwen3-plus'export LLM_CODER_MODEL_NAME='qwen3-coder-plus'export LLM_GENERIC_MODEL_API_ENDPOINT='your-api-endpoint'export LLM_CODER_MODEL_API_ENDPOINT='your-api-endpoint'export LLM_GENERIC_MODEL_API_KEY='your-api-key'export LLM_CODER_MODEL_API_KEY='your-api-key'# Start all services
docker-compose up -d
# Check service health
docker-compose ps

Option 2: Local Development

# Install dependencies
pnpm install
# Start Redis
docker run --rm -p 6379:6379 redis:7-alpine
# Start Kubernetes cluster (required for code execution)cd deploy
docker-compose up kind -d
# Wait for cluster to be ready
docker-compose exec kind sh -c "kubectl wait --for=condition=Ready nodes --all --timeout=300s"# Verify cluster is running
docker-compose exec kind sh -c "kubectl get nodes"# Start Tooling Service (in new terminal)cd apps/mcp-server
uv venv --python 3.12 --seed .venv
source .venv/bin/activate
uv sync
uv run uvicorn mcp_server.main:api --reload --port 8080
# Start AI Agent (in new terminal)cd apps/agent
uv venv --python 3.12 --seed .venv
source .venv/bin/activate
uv sync
export LLM_GENERIC_MODEL_API_KEY='your-api-key-here'export LLM_CODER_MODEL_API_KEY='your-api-key-here'
uv run nat serve --config_file configs/config.yaml
# Start Web App (in new terminal)cd apps/web
pnpm dev
# Optional: Start development tools (in new terminal)cd deploy
# Start Redis Insight for database visualization
docker-compose up redis-insight -d
# Start MCP Inspector for tool debugging
docker-compose up mcp-inspector -d

🌐 Access Points

Service	URL	Description
Web Application	http://localhost:3000	Main user interface
Agent API	http://localhost:8000/docs	AI Agent API documentation
Tooling Service	http://localhost:8080/docs	MCP Server API documentation
Redis Insight	http://localhost:5540	Database visualization (dev only)
MCP Inspector	http://localhost:6274	Tool debugging (dev only)

💡 Development Tools: Redis Insight and MCP Inspector are available via Docker Compose for debugging and monitoring during development.

🎮 Usage Examples

✅ Basic Strategy Request

"Create a moving average crossover strategy for AAPL with 20-day and 50-day periods,
backtest it for the last 2 years, and show me the performance metrics"

🚧 Advanced Technical Analysis (In Roadmap)

"Build an RSI mean reversion strategy for SPY with:
- RSI oversold at 30, overbought at 70
- Bollinger Bands for entry confirmation
- 2% stop loss, 4% take profit
- Backtest on 5-minute data for the last 3 months"

🚧 Multi-Asset Strategy (In Roadmap)

"Create a pairs trading strategy between INTC and AMD using:
- 60-day rolling correlation
- Z-score entry/exit signals
- Risk management with position sizing
- Show me the equity curve and drawdown analysis"

📊 Features Deep Dive

🤖 AI Agent Capabilities

Natural Language Understanding: Interprets complex trading requirements
Multi-Step Planning: Breaks down complex strategies into executable steps
Tool Orchestration: Seamlessly coordinates data fetching, analysis, and execution
Streaming Feedback: Real-time progress updates during strategy development
Error Recovery: Intelligent handling of failures with retry mechanisms

📈 Market Data & Analysis

Real-Time Data: Live market feeds via Yahoo Finance API
Historical Data: Comprehensive historical datasets with multiple timeframes
Technical Indicators: 50+ built-in indicators (RSI, MACD, Bollinger Bands, etc.)
Data Caching: Intelligent Redis-based caching for performance
Data Validation: Automatic data quality checks and preprocessing

🔬 Backtesting Engine

Isolated Execution: Secure, sandboxed code execution environments
Multiple Timeframes: Support for tick, minute, hourly, daily data
Realistic Simulation: Slippage, commissions, and market impact modeling
Performance Metrics: 20+ comprehensive performance indicators
Risk Analysis: Drawdown, Sharpe ratio, maximum adverse excursion

📊 Visualization & Analytics

Interactive Charts: Professional candlestick charts with technical overlays
Performance Dashboards: Real-time KPI monitoring and analysis
Trade Analysis: Detailed trade-by-trade breakdown with entry/exit points
Signal Visualization: Clear buy/sell signal overlays on price charts
Export Capabilities: PDF reports and CSV data export

🛠️ Development

Project Structure

AgentQuant/
├── apps/
│ ├── web/ # Next.js frontend application
│ ├── agent/ # AI agent service (Python)
│ └── mcp-server/ # Tooling service (FastAPI)
├── deploy/ # Docker and Kubernetes configurations
├── assets/ # Static assets (images, demos)
└── LICENSE # AGPL-3.0 license

🛠️ Available Scripts

Monorepo Commands

pnpm build # Build all applications
pnpm dev # Start all services in development mode
pnpm lint # Run linting across all packages
pnpm format # Format code with Prettier
pnpm test# Run test suites

Individual Service Commands

# Web Applicationcd apps/web && pnpm dev
# Tooling Servicecd apps/mcp-server && uv run uvicorn mcp_server.main:api --reload --port 8080
# AI Agentcd apps/agent && uv run nat serve --config_file configs/config.yaml

🔧 Adding New Features

Feature Type	Location	Description
Technical Indicators	`apps/mcp-server/mcp_server/`	Add new trading indicators
UI Components	`apps/web/components/`	Create React components
Agent Tools	`apps/agent/src/agent/`	Extend AI capabilities
API Endpoints	`apps/mcp-server/mcp_server/`	Add new API routes

🚀 Deployment

Production Environments

Environment	Use Case	Description
Docker Compose	Development & Production	Single-server deployment for all use cases
Kind Cluster	Development	Local Kubernetes cluster for testing

Environment Configuration

Variable	Description	Required
`LLM_GENERIC_MODEL_NAME`	Model name for general tasks	✅
`LLM_CODER_MODEL_NAME`	Model name for code generation	✅
`LLM_GENERIC_MODEL_API_ENDPOINT`	API endpoint for generic model	✅
`LLM_CODER_MODEL_API_ENDPOINT`	API endpoint for coder model	✅
`LLM_GENERIC_MODEL_API_KEY`	API key for generic model	✅
`LLM_CODER_MODEL_API_KEY`	API key for coder model	✅
`REDIS_URL`	Redis connection string	✅
`K8S_CONFIG_FILE`	Kubernetes config path	Optional

📚 API Reference

Core Endpoints

Endpoint	Method	Description
`/api/chat`	POST	Submit strategy requests
`/api/data/[storageKey]`	GET	Retrieve market data
`/api/result/[taskId]`	GET	Get backtest results
`/health`	GET	Service health check

MCP Tools

Tool	Description
`task_register`	Register new strategy tasks
`quick_preview`	Preview strategy parameters
`yh_query_save`	Fetch and cache market data
`code_generator`	Generate trading strategy code
`code_executor`	Execute backtest code

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

Development Setup

Fork the repository
Create a feature branch: git checkout -b feature/amazing-feature
Make your changes and add tests
Run the test suite: pnpm test
Commit your changes: git commit -m 'Add amazing feature'
Push to the branch: git push origin feature/amazing-feature
Open a Pull Request

📄 License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0) - see the LICENSE file for details.

🙏 Acknowledgments

LangChain for AI orchestration framework
NVIDIA NAT for agent infrastructure
Alibaba Qwen for performant Large Language Model
FastAPI for high-performance API framework
Next.js for modern React framework
Lightweight Charts for financial visualizations
Redis for high-performance caching

📞 Support

🌐 Official Website: nlpquant.ai - Join our waitlist for early access
📖 Documentation: Wiki
🐛 Bug Reports: Issues
💬 Discussions: GitHub Discussions
📧 Email: support@nlpquant.ai

Built with ❤️ by the NLPQuant Team

Website • GitHub • Documentation

Name		Name	Last commit message	Last commit date
Latest commit History 97 Commits
.idea		.idea
.vscode		.vscode
apps		apps
assets		assets
deploy		deploy
.gitignore		.gitignore
.gitmessage		.gitmessage
.mise.toml		.mise.toml
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml
turbo.json		turbo.json

Uh oh!

Folders and files

Latest commit

History

Repository files navigation