guy-hartstein

company-research-agent: Build Source-Backed Company Briefings

Company Research Agent turns a company name into a structured research briefing using specialized agents, web search, and language models. It is suited to teams that want an interactive, self-hosted starting point for company diligence and report generation.

PythonApache-2.0AI AgentsMulti Agent SystemsWeb App
View on GitHub Compare
Stars
2.3k
Forks
322
Stars, last 7 days
+9

Our take

Health score

43/100

How it is scored
Activity
56
Community
31
Issues
49
Pull requests
35

+9 stars in the last 7 days

  • Slower to merge pull requests than 89% of the projects we track

Overview

Company Research Agent is a web application for researching a company and assembling findings into a structured, source-backed briefing. Its multi-agent workflow gathers and curates material on the company, its industry, financial context, and recent news, then synthesizes and edits the report.

The project is useful when a researcher wants to accelerate initial diligence rather than gather and organize sources manually. It combines a React interface with a FastAPI backend, streams job progress to the browser, and can export completed reports as PDFs. Its output is a research aid, not a substitute for checking sources or making investment decisions.

Key Features

  • Runs a research graph with analyzer, collector, curator, briefing, and editing stages.
  • Uses Tavily search and relevance scoring to find and curate source material.
  • Uses Gemini for briefing synthesis and OpenAI models for research tasks and final report editing.
  • Accepts company details such as name, website, industry, and headquarters location.
  • Streams research progress and completion events to the web interface.
  • Exports report Markdown to PDF.
  • Can persist jobs and reports in MongoDB; persistence is optional.

Use Cases

  • Analysts can create a first-pass company and industry briefing before deeper diligence.
  • Consultants can gather source material and recent news to prepare for client research.
  • Founders or business teams can assemble a structured overview of a competitor or potential partner.
  • Developers can use the application as a reference for building multi-agent research workflows with a web UI and progress streaming.

Getting Started

The project requires Python 3.11 or later, Node.js 18 or later, and backend API keys for Gemini and OpenAI. For a local setup, install backend dependencies with:

uv venv .venv
uv pip install -r requirements.txt

Configure the environment keys and frontend, then start the API and UI as described in the README. Docker Compose is also documented there.

Considerations

  • Running research requires configured Gemini and OpenAI API keys, and Tavily access through a per-session key or an optional backend fallback key. These external services may have their own usage costs and limits.
  • Research quality depends on retrieved sources and model output. Verify important claims against the cited material, especially for financial or investment-related decisions.
  • MongoDB is optional, but without it jobs and reports are not configured for persistent storage.
  • Google Maps credentials are only needed for the optional location autocomplete feature.

Found this useful?

Share it with someone who would like company-research-agent.