Dexter — Autonomous Agent for Deep Financial Research

Introduction

Dexter is an open-source autonomous agent designed for deep financial research. It chains together multiple LLM-powered analysis steps — from data gathering to report writing — to produce structured investment research that would normally take an analyst hours.

What Dexter Does

• Autonomously researches companies using public financial data and filings
• Chains multi-step analysis: data retrieval, financial modeling, comparison, and report generation
• Produces structured research reports with quantitative backing
• Supports configurable research workflows and custom analysis templates
• Integrates with multiple LLM providers for the reasoning layer

Architecture Overview

Dexter is a TypeScript application built around a directed acyclic graph of research steps. Each step is an autonomous agent node that performs a specific task (fetch SEC filings, analyze revenue trends, compare competitors) and passes structured output to downstream nodes. The orchestrator manages execution order, retries, and context sharing. A web UI lets users configure research targets and review generated reports.

Self-Hosting & Configuration

• Clone the repository and install dependencies with npm
• Configure your LLM provider API keys in the environment file
• Set up financial data source credentials (SEC EDGAR is free)
• Customize research templates in the configuration directory
• Run the web interface locally on the default port

Key Features

• Multi-step autonomous research pipeline, not single-prompt generation
• Structured output with citations and data references
• Configurable analysis templates for different research types
• Support for multiple LLM backends (OpenAI, Anthropic, local models)
• Web UI for configuring research targets and reviewing reports

Comparison with Similar Tools

• Bloomberg Terminal — professional financial data platform; Dexter is open-source and focused on automated analysis
• FinGPT — fine-tuned financial LLM; Dexter is an agent workflow, not a single model
• Langchain Financial Agents — generic agent framework; Dexter is purpose-built for investment research
• AlphaVantage + GPT — manual API + prompt combo; Dexter automates the full pipeline

FAQ

Q: Does Dexter provide investment advice? A: No. It generates research reports as a tool for analysts. All output should be reviewed by a human before making investment decisions.

Q: What financial data sources does it use? A: It primarily uses public data from SEC EDGAR, Yahoo Finance, and similar free sources.

Q: Can I use local LLMs instead of cloud APIs? A: Yes. Dexter supports any OpenAI-compatible API endpoint, including local servers.

Q: Is it suitable for real-time trading? A: No. Dexter is designed for deep research, not high-frequency or real-time trading.

Sources

• https://github.com/virattt/dexter