PortMan

WallStreet Agents democratizes institutional-grade investment research using 10 AI agents built with Claude Opus 4.6. Enter a ticker and agents analyze in parallel — Fundamental, Technical, Sentiment, Macro, Sector, Risk, ESG, and Quant agents gather real-time data via a custom MCP Finance Server, then a Portfolio Manager sizes positions, and an investment committee conducts a 3-round debate using the Agent Teams SDK to produce a BUY/SELL/HOLD memo. Human-in-the-loop: users chat with agents to challenge findings, inject thesis with PDFs to trigger analysis cascades, and add custom Skills for firm-specific methodologies. An MCP Reports Server exposes completed reports to Claude Desktop. A companion AgentCoder platform (7-agent team) autonomously builds tools to connect to any API or database with security auditing and encrypted credential storage. Built with Claude Code SDK, Agent Teams, custom MCP servers, Skills, Hooks, and per-agent effort controls.