Minsky

A risk-focused financial research agent for traditional markets and crypto.

Overview

Minsky is an autonomous research agent that prioritizes robustness over prediction. It analyzes markets through the lens of risk, volatility, and tail events—challenging fragile narratives rather than reinforcing them.

Philosophy:

• Fat tails and black swans matter more than normal distributions
• What can go wrong eventually will—focus on survivability
• Consensus narratives often mask fragility
• Volatility is information, not just noise

Coverage:

• Traditional finance: equities, bonds, commodities, derivatives
• Crypto: tokens, DeFi protocols, stablecoins

Setup

Prerequisites

• Node.js (v18+)
• xAI API key
• Perplexity API key

Installation

git clone https://github.com/virattt/minsky.git
cd minsky
npm install

Configuration

cp env.example .env

Edit .env and add your API keys:

XAI_API_KEY=your-xai-api-key
PERPLEXITY_API_KEY=your-perplexity-api-key

Usage

npm run dev

Example Queries

• "What are the tail risks for NVIDIA at current valuations?"
• "Analyze Bitcoin's correlation with risk assets during stress events"
• "What hidden leverage exists in the current market structure?"
• "Compare the liquidity profile of SOL vs ETH"
• "What could break the soft landing narrative?"

Architecture

Built on LangGraph with a ReAct agent pattern:

1. Research: Fetches real-time data via Perplexity
2. Analysis: Synthesizes information with risk-first framing
3. Response: Delivers insights focused on robustness and tail risks

License

MIT