A hedge fund in a box, with an actual kill switch

What it does AI Market Maker is an open-source trading stack that simulates a small hedge fund’s workflow: specialist agents (market scan, technical analysis, statistical alpha, sentiment) generate signals, a portfolio manager sizes positions, and a hard “Risk Guard” can veto any trade before execution. It backtests against buy-and-hold benchmarks, paper-trades on Binance Testnet, and surfaces everything through a Next.js dashboard.

The interesting bit Most trading bots log “risk checks passed.” This one gives Risk Guard actual veto power — the LangGraph workflow halts if the safety layer objects. The README is unusually self-aware about backtest limitations, warning that “a single profitable backtest is not proof of edge.” That quant-grade caution is rare in open-source trading projects.

Key highlights

• LangGraph orchestration with standardized Input → Process → Output → Feedback agent contracts
• Hard Risk Guard veto layer, not advisory logging
• Built-in benchmark comparison (excess return vs. buy-and-hold) on every backtest
• OpenClaw-ready packaging with SKILL.md + manifest.json for external runner integration
• Paper trading on Binance Testnet; Hyperliquid adapter in dry-run via OMS layer
• Futu OpenD adapter for HK/US stock data (requires separate gateway)

Caveats

• Currently limited to paper/dry-run execution; full position lifecycle and leverage config are near-term goals
• Requires TA-Lib C library installation, which the README admits can be finicky
• LLM nodes are optional but the system clearly expects OpenAI integration for full agentic behavior

Verdict Worth exploring if you’re building (or skeptical of) agentic trading systems and want a reference architecture with real risk controls. Skip it if you need production live trading today — this is deliberately a research-and-paper-trading platform.