guy-hartstein/company-research-agent

An agentic company research tool powered by LangGraph and Tavily that conducts deep diligence on companies using a multi-agent framework. It leverages Google's Gemini 2.5 Flash and OpenAI's GPT-5.1 on the backend for inference.

What it solves

It automates the process of conducting deep company research, replacing manual searching and synthesis with a multi-agent pipeline that generates comprehensive research reports.

How it works

The system uses a sequential pipeline of specialized AI agents to gather and process data:

  1. Research Phase: Specialized nodes (CompanyAnalyzer, IndustryAnalyzer, FinancialAnalyst, and NewsScanner) gather data from websites, news, financial reports, and industry analyses.
  2. Curation Phase: A Collector aggregates the data, and a Curator uses Tavily's relevance scoring to filter out low-quality content.
  3. Synthesis Phase: A Briefing node uses Gemini 2.5 Flash to summarize large volumes of data into category-specific briefings.
  4. Finalization Phase: An Editor node uses GPT-5.1 to deduplicate content and format the final report into a structured Markdown document.

The backend is built with FastAPI and supports asynchronous processing, while the frontend is a React application that polls for the report's completion status.