Dexter is an open-source autonomous AI agent designed for deep financial research and analysis. It takes complex financial questions and turns them into clear, step-by-step research plans that are autonomously executed with self-validation. Features include intelligent task planning that decomposes complex queries into structured research steps, autonomous tool selection and execution for data gathering, self-validation that iterates until tasks complete, real-time financial data access including income statements, balance sheets, and cash flow data, and built-in safety features including loop detection and step limits. The project includes an interactive CLI mode, development watch mode, evaluation suite with LangSmith integration using LLM-as-judge scoring, debug scratchpad logging in JSONL format, and WhatsApp gateway integration for mobile access to financial research. Built entirely in TypeScript running on Bun runtime, it requires an OpenAI API key and Financial Datasets API key, with optional Exa API key for web search. The latest release v2026.3.25 reflects active development with 399 commits and 21 contributors.