An AI trading desk built as a team of sub-agents (Claude Code and Robinhood MCP) (github.com)

A new AI trading desk, designed as a multi-agent system within Claude Code, connects to a Robinhood Agentic account, enabling users to conduct stock research while retaining full control over trading decisions. This setup features specialized sub-agents focused on fundamental analysis, technical analysis, macro news, and risk management, which collaboratively evaluate potential investments. Importantly, no trades can be executed without user approval, ensuring a human-in-the-loop model that prioritizes security and informed decision-making. This innovative architecture is significant for the AI/ML community as it represents a structured approach to automating stock research without independently executing trades, effectively mitigating risks associated with algorithmic trading. The system is safeguarded by prompt-injection defenses, as agents treat external news data as untrusted. It operates with a clear set of guardrails, including position caps and read-only research that limits interactions to essential queries. Although currently in beta and limited to U.S. equities within a designated budget, this AI trading desk could serve as a reference for future developments in automated trading systems that blend AI capabilities with prudent human oversight.