Summary

Nunchi-trade/agent-cli provides a comprehensive, autonomous trading stack for Hyperliquid, enabling sophisticated market making, arbitrage, and directional strategies. Targeting technically savvy users, it offers an advanced autonomous trading solution with integrated LLM capabilities and a self-improvement loop, aiming to automate and optimize crypto trading performance.

How It Works

The system employs a modular architecture featuring 14 distinct strategies built upon a common quoting engine pipeline. Core components include the APEX orchestrator for multi-position management, Radar for opportunity screening, Pulse for momentum detection, and Guard for dynamic trailing stops. A novel REFLECT engine performs nightly performance reviews and auto-adjusts parameters, while an optional claude_agent integrates LLMs for decision-making.

Quick Start & Requirements

• Installation involves cloning the repository and running bash scripts/bootstrap.sh to set up a virtual environment and dependencies.
• Essential setup requires configuring the HL_PRIVATE_KEY environment variable for wallet access. LLM integration necessitates API keys for providers like OpenAI, Anthropic, or Gemini. Testnet usage requires HL_TESTNET=true.
• Official documentation is available at docs.nunchi.trade.

Highlighted Details

• 14 built-in strategies covering market making, arbitrage, directional signals, and infrastructure roles.
• Full autonomous stack: APEX orchestrator, Radar, Pulse, Guard, and REFLECT self-improvement loop.
• LLM integration (claude_agent) for AI-driven trading decisions via Gemini, Claude, or OpenAI.
• MCP server exposes 16 trading tools for AI agent integration (stdio/SSE).
• Deployment options include headless APEX (deterministic) or OpenClaw agent (conversational AI via Telegram).
• Production safety features: Exchange-Level Stop Loss Sync, Clearinghouse Reconciliation, Risk Guardian.
• Support for YEX Yield Markets (VXX-USDYP, US3M-USDYP, BTCSWP-USDYP).

Maintenance & Community

• Community support is available via Discord (discord.gg/nunchi) and X (@nunchi).
• Comprehensive documentation is hosted at docs.nunchi.trade.

Licensing & Compatibility

• The project is released under the permissive MIT License, allowing for commercial use and integration into closed-source applications.

Limitations & Caveats

• Requires sensitive HL_PRIVATE_KEY configuration. LLM features incur external API costs. The claude_agent is noted as experimental. Some strategies depend on the quoting_engine module.