virattt

dexter: Research Financial Questions with an Autonomous Agent

Dexter turns complex financial questions into research plans, gathers market data with tools, and checks its work before responding. It is aimed at developers and researchers who want an extensible, command-line financial research agent, not a source of investment advice.

View on GitHub Compare
Stars
27.6k
Forks
3.4k
Stars, last 30 days
+165

Our take

Health score

37/100

How it is scored
Activity
68
Community
39
Issues
21
Pull requests
20

+165 stars in the last 30 days

  • Releases more often than 82% of the projects we track

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

27,837

Forks

3,434

Change since Nov 14, 2026: +3 stars·0 forks

Overview

Dexter is a TypeScript agent for researching financial questions. It breaks a query into steps, uses tools to collect financial data, and reviews its results to produce a data-backed response.

It is suited to developers and researchers who want to experiment with automated financial research or adapt an agent to their own workflows. It is not intended for making trading or investment decisions.

Key Features

  • Plans multi-step research from complex questions.
  • Chooses tools to gather financial information.
  • Retrieves income statements, balance sheets, and cash flow statements using financial data services.
  • Validates and refines its work before returning an answer.
  • Uses loop detection and step limits to help prevent runaway execution.
  • Saves tool calls and results in JSONL scratchpad files for inspection.
  • Offers an optional WhatsApp gateway for sending questions and receiving responses.
  • Includes an evaluation runner for financial questions, with LangSmith tracking and LLM-based scoring.

Use Cases

  • Developers prototyping an agent that can break financial research questions into executable steps.
  • Analysts exploring company financial statements through an interactive research workflow.
  • Researchers inspecting tool calls and logged results to understand how an agent reached an answer.
  • Teams evaluating an agent's responses against a dataset of financial questions.
  • Users who want to submit research queries through WhatsApp rather than the terminal.

Getting Started

Install Bun and configure the required OpenAI and Financial Datasets API keys. Then:

git clone https://github.com/virattt/dexter.git
cd dexter
bun install
bun start

See the repository README for environment setup, optional providers, evaluation, and gateway instructions.

Alternatives

  • FinceptTerminal: FinceptTerminal is a desktop research suite with analytics and paper trading, while Dexter is an extensible CLI agent focused on answering financial questions.
ProjectLanguageLicenseStarsStatus
dexterTypeScript–27.6kNot checked yet
FinceptTerminalC++–32.3kActive

Considerations

  • Requires Bun and API keys for an LLM provider and Financial Datasets. Exa is optional for web search.
  • The README warns that outputs may be incorrect, incomplete, or out of date. The project is for educational and informational use, not financial, investment, tax, or legal advice.
  • The repository input does not specify a license.

Found this useful?

Share it with someone who would like dexter.

Comparisons