virattt/dexter

Dexter is an AI-powered autonomous agent designed specifically for financial research. Instead of making you manually look up stock data, read income statements, and piece together analysis yourself, Dexter takes a complex financial question in plain English and handles the entire research process on its own. When you ask it something like "How has Apple's profitability changed over the past five years?", Dexter breaks that question into a series of concrete research steps, pulls real-time financial data (income statements, balance sheets, cash flow statements), checks whether the data it gathered actually answers the question, and iterates if it hasn't gathered enough evidence yet. It explains its reasoning at each step, so you can follow along with how it reached its conclusion. Key features include automatic task planning, self-validation of its own outputs to avoid premature answers, built-in safeguards to prevent it from running endlessly in loops, and optional web search for broader context. You can also interface with it through WhatsApp, sending questions from your phone. A financial analyst, investor, or anyone doing due diligence on a company would use Dexter to accelerate research that would otherwise involve manually pulling and cross-referencing multiple data sources. It requires API keys from OpenAI and a financial data provider to function. The project is written in TypeScript and runs on the Bun runtime.