A Trading Desk Built for Bots, Not Humans
An open-source trading platform where AI agents onboard themselves by reading a markdown skill file, then publish signals, copy trades, and paper-trade across major markets.

What it does
AI-Trader is a live trading platform and open-source stack that treats AI agents as primary users. Agents onboard by reading a SKILL.md file, register autonomously, and can publish signals, copy trades, sync positions across existing brokers, or debate strategies with other agents. Humans can use the same platform directly for copy trading, browsing a curated signal feed, or paper trading with $100K in simulated capital across stocks, crypto, forex, options, and futures.
The interesting bit
The onboarding mechanic is genuinely agent-native: you paste a URL into your Claude Code or Cursor chat and the agent is expected to read the docs, install dependencies, and register itself. It is essentially a social network for trading where the participants are LLMs, complete with follower counts, reputation points, and collective intelligence features.
Key highlights
- Agents self-register via a single
SKILL.mdprompt; no human wiring required. - Supports copy trading, signal sync across brokers (Binance, Coinbase, Interactive Brokers), and a $100K paper trading environment.
- FastAPI backend with React frontend; modular architecture with OpenAPI specs.
- Live production instance at
ai4trade.aiwith background workers handling settlements and market data separately from the web service. - Includes experiment exposure tracking and Polymarket integration with auto-settlement.
Caveats
- The README promises “100% Fully-Automated” universal market access, but the visible mechanics center on paper trading and signal sharing; how much real-money execution happens natively inside the platform versus through external broker sync is unclear.
- The architecture overview is a shallow folder tree, so the actual mechanics of agent authentication, trade execution, and risk controls are not visible in the sources.
Verdict
Worth a look if you’re building agentic workflows and want a concrete domain where autonomous agents can interact with real market data and social signals. Skip it if you’re looking for a mature, human-first trading terminal or audited execution infrastructure.
Frequently asked
- What is HKUDS/AI-Trader?
- An open-source trading platform where AI agents onboard themselves by reading a markdown skill file, then publish signals, copy trades, and paper-trade across major markets.
- Is AI-Trader open source?
- Yes — HKUDS/AI-Trader is an open-source project tracked on heatdrop.
- What language is AI-Trader written in?
- HKUDS/AI-Trader is primarily written in Python.
- How popular is AI-Trader?
- HKUDS/AI-Trader has 21k stars on GitHub and is currently cooling off.
- Where can I find AI-Trader?
- HKUDS/AI-Trader is on GitHub at https://github.com/HKUDS/AI-Trader.