Dexter just hit 18K stars — it's Claude Code for finance

Imagine asking an AI "How has Netflix's revenue per user changed over the past 5 years?" and getting back a verified, sourced answer with real financial data — in under a minute. That's what Dexter does. It's a free, open-source AI agent that does deep financial research autonomously, like having a Wall Street analyst in your terminal.

Created by Virat Singh, a developer who quit his job to build it, Dexter has exploded to 18,800 GitHub stars and 2,300+ forks. Its creator describes it as "Claude Code, but for finance" — and the benchmarks back that up: 92% faster, 26% cheaper, and 31% more accurate than using general-purpose AI tools for financial questions.

It thinks, plans, and checks its own work

What makes Dexter different from just asking ChatGPT about stocks is its four-agent architecture — four specialized AI roles that work together on every question:

This self-validation loop is the key difference. General-purpose AI tools often sound confident while getting financial numbers wrong. Dexter catches its own mistakes before you see them.

What you can actually ask it

Dexter handles the kinds of questions that would normally take an analyst 10-30 minutes of digging through filings and spreadsheets:

• "Did TJX beat or miss its Q4 pre-tax margin guidance?" — pulls the actual guidance, actual results, and calculates the difference
• "How has AMD's revenue guidance range changed over the past 4 quarters?" — gathers data from multiple earnings calls
• "How did US Steel address its merger with Nippon Steel?" — reads SEC filings and extracts the key points
• "Compare Apple and Microsoft's free cash flow trends" — pulls both companies' financials and runs the comparison

It pulls data from income statements, balance sheets, cash flow reports, SEC filings (10-K annual reports, 10-Q quarterly reports, 8-K event reports), insider trades, earnings data, and analyst estimates.

Works with any AI model — including free local ones

Dexter isn't locked to one AI provider. It works with OpenAI, Anthropic (Claude), Google Gemini, xAI (Grok), OpenRouter, and — critically — Ollama, which lets you run it entirely on your own computer with no API costs at all. For Anthropic users, it supports prompt caching (a way to reuse previous context) that can cut costs by up to 90%.

There's even a WhatsApp integration — link your phone and chat with Dexter like texting a friend who happens to be a financial analyst.

Get it running in 3 minutes

You'll need Bun (a fast JavaScript runtime) installed, plus at least one AI API key:

git clone https://github.com/virattt/dexter.git
cd dexter
bun install
cp env.example .env
# Add your API key to .env, then:
bun start

For a free API key for financial data, sign up at financialdatasets.ai (also created by Virat). To use it with a completely local AI model through Ollama, no API key is needed at all.

The transparency factor

Every research session is logged in detail to .dexter/scratchpad/ files — you can see exactly what tools were called, what data was fetched, and how the AI reasoned through each step. This is a major advantage over closed platforms like Perplexity Finance, where you can't verify the reasoning chain.

The project ships weekly updates (the latest, v2026.3.25, added a stock screener and merged financial search tools) and has an active community on Discord. Gergely Orosz, author of The Pragmatic Engineer (one of tech's most-read newsletters), called it "amazing" and praised the innovation happening in command-line tools.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News