Chao Huang launches AI‑Trader 2.0

Software agents are programs that can read data, make decisions, and call tools on their own. Chao Huang has now turned that idea into a public trading platform called AI‑Trader 2.0, with code released on GitHub. (github.com) The repository says AI‑Trader is “100% fully-automated” and “agent-native,” meaning it is built for machine agents first rather than for human traders clicking through dashboards. The project sits in the Hong Kong University of Science and Data Science Lab GitHub organization, where Huang’s lab account lists AI‑Trader among 87 public repositories. (github.com 1) (github.com 2) The current README says the platform supports major agent clients including OpenClaw, nanobot, Claude Code, Codex, and Cursor. It also points users to a live service at ai4trade.ai and says an agent can join by reading a skill file and registering itself. (github.com 1) (github.com 2) In plain terms, the system is trying to give trading bots their own workplace. One agent can publish signals, another can follow them through copy trading, and others can pull market data or Polymarket information through separate skill files and application programming interfaces. (github.com 1) (github.com 2) The codebase shows a shift from a chat-style assistant toward a network of narrow specialists. The project documentation breaks the platform into functions such as marketplace selling, trade syncing, copy trading, heartbeat monitoring, market intelligence, and Polymarket data access. (github.com) (github.com) That design mirrors a broader pattern in agent research. A separate TradingAgents framework describes a multi-agent trading setup with distinct roles for analysts, researchers, traders, risk managers, and fund managers rather than one general model doing everything. (tradingagents-ai.github.io) AI‑Trader’s recent update log shows the project moving quickly in March and April 2026. The README lists Polymarket paper trading on March 3, a dashboard launch on March 21, codebase streamlining on April 9, and production stability hardening on April 10. (github.com) The GitHub page also shows strong early developer attention. The HKUDS organization page listed AI‑Trader at about 12,737 stars and 2,144 forks when it was crawled, with the repository updated “yesterday.” (github.com) The project’s own documentation describes a marketplace for trading signals, data feeds, and artificial intelligence models, with email-based registration and token-based agent access. It also says new users receive 100 welcome points and can publish signals or follow providers through copy trading. (github.com) (github.com) For now, the launch puts Huang’s bet in public view: if human traders have Bloomberg terminals and broker apps, autonomous agents may get skill files, shared signals, and a platform built to let them trade with each other directly. (github.com)