Dexter Proves AI Agents Can Master Financial Data

In this repo deep dive, we look at Dexter by virattt: an open-source finance agent for deep financial research. Dexter plans research tasks, calls finance-specific tools, reads filings, searches the web, keeps scratchpad logs, supports memory, includes evals, and can even run through WhatsApp.

The bigger idea: when agents can parse filings, financial statements, news, and market history quickly, the advantage shifts from finding information to asking sharper questions.

Repo: https://github.com/virattt/dexter

**Subscribe to Build Things With AI**

Chapters
0:00 Financial Research Is Becoming Agentic
0:26 Claude Code Energy, Finance Domain
0:52 Finance Data Is Built In
1:18 Not A Trading Bot
1:41 The Agent Leaves A Paper Trail
2:05 24.1K Stars, TypeScript Core
2:35 Terminal App, Finance Tools, Memory, Gateway
3:01 Plan, Use Tools, Check, Continue
3:28 Natural Language Routes The Data
3:55 Numbers Need Filing Context
4:22 Structured Data Meets Live Context
4:47 Research Context Carries Forward
5:16 Finance Agents Need Scoreboards
5:46 The Agent Can Live In WhatsApp
6:08 Bun App, API Keys, Terminal UI
6:36 Recent Work Polishes The Agent UI
7:03 The Scarce Skill Becomes The Question
7:28 Better Than A Generic Agent Wrapper
7:54 Use It As Research Infrastructure
8:19 A Vertical Agent For Market Research

Видео Dexter Proves AI Agents Can Master Financial Data канала Build Things With AI