Skip to content
/ quant-connect-mcp-lambda Public

Stateless MCP Server in AWS Lambda for AI assistants to interact with QuantConnect's algorithmic trading platform.

License

MIT license
2 stars 0 forks Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

fschaeffler/quant-connect-mcp-lambda

BranchesTags

Repository files navigation

QuantConnect MCP Server

Model Context Protocol server for AI assistants to interact with QuantConnect's algorithmic trading platform.

Overview

Bridge between AI assistants and QuantConnect platform for:

  • Algorithm development and management
  • Backtesting and live trading deployment
  • Data management and research tools
  • Project collaboration and optimization

Architecture

The project follows a monorepo structure with multiple specialized packages:

aws-lambda-mcp/
├── libs/                           # Core libraries
│   ├── quant-connect-mcp/         # Main MCP server implementation (67 tools)
│   ├── quant-connect-types/       # TypeScript types from QuantConnect API
│   ├── quant-connect-client/      # HTTP client for QuantConnect API
│   └── middy/                     # AWS Lambda middleware integration
├── services/                      # Deployable services
│   ├── quant-connect/            # Main Lambda service
│   └── quant-connect-local/      # Local development service
├── infra/                        # AWS CDK infrastructure
└── e2e/                          # End-to-end testing

Core Components

  • MCP Server: Model Context Protocol implementation with 67 tools organized by category
  • QuantConnect Client: HTTP client with authentication, rate limiting, and error handling
  • AWS Lambda: Serverless deployment with API Gateway, Secrets Manager integration
  • Type Safety: Auto-generated TypeScript types with Zod validation
  • Local Development: HTTP server for development without AWS infrastructure

Note: Uses modified OpenAPI spec to fix validation issues in the official specification.

vs Official QuantConnect MCP Server

Alternative to official MCP server with key differences:

Architecture

Feature This Implementation Official Server
Language TypeScript/Node.js Python
Deployment AWS Lambda + API Gateway Docker container
Runtime Serverless Container-based
Scaling Auto-scaling Lambda Manual container scaling
Infrastructure AWS CDK (Infrastructure as Code) Docker Compose/Kubernetes

Tools

Category This Implementation Official Server
Total Tools 67 tools 64 tools
Project Management ✅ Full support ✅ Full support
File Operations ✅ Full support ✅ Full support + patch_file
Backtesting ✅ Full support ✅ Full support
Live Trading ✅ Full support ✅ Full support
AI Tools ✅ Full support ✅ Full support
Optimization ✅ Full support ✅ Full support
Object Store ✅ Full support ✅ Full support

Key Differences

Missing Features:

  • patch_file tool (git-style patches)
  • authorize_connection tool (two-step live auth)

Unique Features:

  • AWS cloud integration (Lambda, API Gateway, Secrets Manager, CloudWatch)
  • Infrastructure as Code (CDK)
  • Full TypeScript type safety with Zod validation
  • Local development server
  • Built-in monitoring and authentication

Development

Aspect This Implementation Official Server
Setup Node.js ecosystem Python ecosystem
Type Safety Full TypeScript + Zod validation Python type hints
Testing Jest + comprehensive test suite Python testing framework
Documentation Extensive MD docs + API reference README + inline docs
Local Development HTTP server + nodemon for dev reloading Python STDIO + inspector
IDE Support Full IntelliSense + autocomplete Python IDE support

Production

Factor This Implementation Official Server
Security AWS IAM + API Gateway auth Container security
Monitoring CloudWatch logs + metrics Container logging
Cost Pay-per-request (serverless) Fixed container costs
Maintenance Managed AWS services Container updates needed
Reliability AWS SLA (99.95%+) Container orchestration dependent
Global Distribution Multi-region AWS deployment Manual geographic deployment

Features

67 tools organized across categories:

Project Management (5 tools)

  • Create, read, update, delete projects
  • List user projects

File Operations (5 tools)

  • Create, read, update, delete algorithm files
  • Update file names and contents

Backtesting (10 tools)

  • Create and manage backtests
  • Read results, charts, orders, insights, reports
  • Historical simulation analytics

Live Trading (12 tools)

  • Deploy algorithms to live trading
  • Monitor portfolio, orders, logs
  • Risk controls and liquidation

AI-Powered Tools (6 tools)

  • Code completion and syntax checking
  • Error message enhancement
  • PEP8 formatting
  • QuantConnect documentation search

Research & Data (8 tools)

  • Object store for research data
  • Parameter optimization
  • Compilation services
  • Account information

Collaboration & Infrastructure (21 tools)

  • Project collaboration tools
  • Node management
  • Live command broadcasting
  • Health monitoring
  • Version information

Quick Start

Prerequisites: Node.js 18+, QuantConnect account

Local Development

# Install dependencies
git clone <repository-url>
cd aws-lambda-mcp
npm ci
npm run build

# Configure credentials
cd services/quant-connect-local
echo "QUANTCONNECT_USER_ID=your-user-id" > .env
echo "QUANTCONNECT_API_TOKEN=your-api-token" >> .env

# Start development server
npm start
# Server runs at http://localhost:55555/mcp

# Test the connection
curl -X POST http://localhost:55555/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

AWS Production Deployment

# Deploy infrastructure (creates empty secrets)
cd infra && npm run deploy

# Populate secrets with actual QuantConnect credentials
aws secretsmanager update-secret \
  --secret-id "quant-connect-mcp/user-id" \
  --secret-string "your-quantconnect-user-id"
aws secretsmanager update-secret \
  --secret-id "quant-connect-mcp/api-token" \
  --secret-string "your-quantconnect-api-token"

After deployment, you'll receive API Gateway endpoint and API key for authentication.

Configuration

Local Development

  • Environment Variables: QUANTCONNECT_USER_ID, QUANTCONNECT_USER_ID in .env file
  • Server: Runs on port 55555 (configurable via PORT)
  • Debug: Enable with DEBUG=* or DEBUG=quant-connect:*

AWS Production

  • API Gateway: REST endpoint with API key authentication
  • Lambda Function: Node.js 22.x runtime, 1024MB memory, 30s timeout
  • Secrets Manager: Encrypted credential storage
  • CloudWatch: Logging and monitoring with 3-day retention

MCP Protocol Support

Enhanced MCP protocol implementation:

  • Full JSON-RPC 2.0 compliance with proper request/response handling
  • Notification support for MCP client handshake (initialize/initialized flow)
  • HTTP transport compatibility with AWS Lambda/API Gateway
  • Error handling with appropriate HTTP status codes
  • 67 tools with comprehensive input/output validation using Zod schemas

AWS Infrastructure

Components

  • API Gateway: REST API with throttling (100 req/sec, 200 burst)
  • Lambda Function: Serverless execution with automatic scaling
  • Secrets Manager: Secure credential storage with rotation support
  • CloudWatch: Structured logging and performance metrics
  • IAM Roles: Minimal permissions (secrets access, logging)

Security Features

  • API key authentication for all requests
  • HTTPS/TLS encryption in transit
  • Encrypted secrets and logs at rest
  • Zod input validation for all tools
  • No sensitive data in logs or responses

Scaling & Performance

  • Automatic scaling: Up to 1000 concurrent Lambda executions
  • Cold start: ~2-3 seconds initial, ~500ms subsequent
  • Typical response: 1-5 seconds
  • Memory usage: 150-300MB typical, up to 500MB for large operations

Development

Monorepo Structure

npm run lint        # Code linting across all packages
npm run build       # Build all packages  
npm test           # Run unit tests
npm run test:e2e   # End-to-end tests (requires deployed server)

Hot Reloading for Development

Enable file watching during development:

# Install nodemon globally
npm install -g nodemon

# Start with automatic restart on changes
cd services/quant-connect-local
nodemon --watch src --ext ts,js --exec "npm start"

Testing Strategy

  • Unit Tests: 100% coverage for all TypeScript files
  • Integration Tests: Component interaction validation
  • E2E Tests: Real server testing with AWS authentication
# Run specific test suites
npm test                    # All unit tests
npm run test:coverage      # With coverage report
cd e2e && npm run test:e2e # End-to-end tests

E2E tests validate:

  • API Gateway authentication with API keys
  • Complete MCP protocol compliance
  • All 67 tools functionality
  • Real QuantConnect API integration

Type Safety & Validation

Generated Types

  • Source: Auto-generated from QuantConnect OpenAPI 2.0.0 specification
  • Validation: Zod schemas for runtime validation
  • Features: Automatic date parsing, null safety, coercion

Tool Validation

All tools include:

  • Input validation: Zod schema validation for parameters
  • Output validation: Structured response validation
  • Error handling: Comprehensive error mapping
  • Annotations: Tool behavior hints (readOnly, destructive, idempotent)

Documentation

Complete Documentation Guide

Architecture Documentation

Each component includes comprehensive documentation covering setup, configuration, usage examples, troubleshooting, and best practices.

Cost Optimization

AWS Pricing (Estimated)

  • Lambda: ~$0.0005 per invocation (1024MB, 30s avg)
  • API Gateway: $3.50 per million API calls
  • Secrets Manager: $0.40 per secret per month
  • CloudWatch: $0.50 per GB logs ingested

Optimization Strategies

  • Right-size Lambda memory allocation
  • Implement request caching for repeated operations
  • Use provisioned concurrency for predictable workloads
  • Monitor and optimize based on CloudWatch metrics

Security

  • AWS Secrets Manager: Encrypted credential storage with automatic rotation
  • API Gateway: Authentication, rate limiting, and request validation
  • Zod Validation: All inputs validated against TypeScript schemas
  • IAM Permissions: Minimal required permissions (secrets read, logging)
  • TLS Encryption: All data encrypted in transit
  • Audit Logging: All requests logged to CloudWatch

Troubleshooting

Common Issues

  1. Authentication Errors

    • Verify QuantConnect credentials are correct
    • Check AWS Secrets Manager configuration (production)
    • Ensure environment variables are set (local development)

  2. Cold Start Timeouts

    • First Lambda invocation may take 2-3 seconds
    • Consider provisioned concurrency for production workloads

  3. Rate Limiting

    • API Gateway enforces 100 req/sec by default
    • Implement exponential backoff for high-volume usage

  4. Tool Validation Errors

    • Check input parameters match expected Zod schemas
    • Review tool documentation for required fields

Debug Mode

Enable verbose logging:

# Local development
DEBUG=* npm start

# Lambda (via environment variables)
LOG_LEVEL=debug

Acknowledgments

This project's Middy MCP integration is inspired by the middy-mcp package by Frédéric Barthelet. While our implementation is independently developed with different architectural choices, we acknowledge the foundational work done by the middy-mcp community in establishing patterns for MCP server integration with AWS Lambda.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

About

Stateless MCP Server in AWS Lambda for AI assistants to interact with QuantConnect's algorithmic trading platform.

Resources

Readme

License

MIT license

Contributing

Contributing
Activity

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages