A Python-based multi-agent research tool that conducts company diligence using a pipeline of research and processing nodes, leveraging Tavily and Gemini for high-context synthesis and GPT-5.1 for final formatting. It provides setup scripts, Docker deployment, and API endpoints for running and retrieving reports.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
A multi-agent tool that generates comprehensive company research reports. The platform uses a pipeline of AI agents to gather, curate, and synthesize information about any company.
How it works
The system employs specialized Research Nodes (CompanyAnalyzer, IndustryAnalyzer, FinancialAnalyst, NewsScanner) and Processing Nodes (Collector, Curator, Briefing, Editor). Gemini 2.5 Flash handles high-context synthesis for category briefings, while GPT-5.1 formats and edits the final report. Relevance scoring is performed by Tavily's search, with a default threshold of 0.4. Backend uses FastAPI with async support; frontend is React, with polling-based job status and REST endpoints.
Getting started
Quick Setup (Recommended)
Clone the repository and run setup.sh:
git clone https://github.com/guy-hartstein/company-research-agent.git
cd company-research-agent
chmod +x setup.sh
./setup.sh
The setup script detects uv usage for faster installation, checks versions, optionally creates a virtual environment, installs dependencies, guides environment variables, and can start both backend and frontend servers. You will need Tavily, Gemini, OpenAI, Google Maps API keys, and an optional MongoDB URI.
Manual Setup
git clone https://github.com/guy-hartstein/company-research-agent.git
cd company-research-agent
Install backend dependencies and, if desired, activate a virtual environment, then install frontend dependencies and set up environment variables as described in the README.
Docker Setup
git clone https://github.com/guy-hartstein/company-research-agent.git
cd company-research-agent
docker compose up --build
Backend API available at http://localhost:8000 and frontend at http://localhost:5174. To stop:
docker compose down
Recent releases
- v2.1.0 (2026-07-04): All Tavily calls run at advanced depth for search and extraction tasks.
- 2.0.1 (2025-12-17): Dependency bumps in UI (glob, js-yaml).
- v2.0.0 (2025-11-18): Frontend/backend handshake modernization, Langchain v1.0.
- v1.1.0 (2025-08-26): Added Tavily Crawl for grounding node.
- v1.0.0 (2025-05-21): Initial release with 1K stars.
Traction
Stars: 2204 Forks: 305 Open issues: 4
Behind the repo
This project integrates Tavily API, Google's Gemini, and OpenAI for backend inference, with a multi-agent research pipeline and a React frontend.
Caveats
License: Apache-2.0. Requires multiple API keys (Tavily, Gemini, OpenAI, Google Maps) and optional MongoDB URI. Implemented as Docker and local setup options; setup includes environment variable configuration.