A multi-agent company due diligence and research tool based on LangGraph + Tavily that automatically generates comprehensive company research reports.
company-research-agent is an automated report generation tool designed for investment due diligence, competitive intelligence, and market research. The project utilizes a multi-agent pipeline architecture featuring four specialized research nodes: CompanyAnalyzer, IndustryAnalyzer, FinancialAnalyst, and NewsScanner, which concurrently gather data from multiple sources including official websites, news articles, and financial reports.
Core Architecture
Agent pipeline flow: Research Nodes → Collector → Curator → Briefing → Editor → Final Report
- Research Nodes: CompanyAnalyzer (core business info), IndustryAnalyzer (market position & trends), FinancialAnalyst (financial metrics & performance), NewsScanner (recent news & updates)
- Collector: Aggregates raw research data from all analyzers
- Curator: Filters content using Tavily relevance scoring (default threshold 0.4), with document deduplication and standardization
- Briefing: Invokes Gemini 2.5 Flash for large-context processing and categorized summary generation
- Editor: Invokes GPT-5.1 for precise Markdown formatting, deduplication, and final report compilation
Key Capabilities
- Multi-source concurrent data collection with AI-driven content filtering
- Dual-model collaboration: Gemini 2.5 Flash for comprehensive summaries, GPT-5.1 for formatted output
- Asynchronous polling architecture for background research progress tracking
- Real-time streaming Markdown output with one-click PDF export
- Frontend integration with Google Maps
Deployment Options
Supports local script installation (setup.sh), Docker Compose deployment, and cloud platforms including AWS Elastic Beanstalk, Heroku, Google Cloud Run, and LangGraph Platform. Backend built with FastAPI (port 8000), frontend with React (Vite) + TypeScript (port 5173, or 5174 in Docker).
API Endpoints
| Method | Path | Description |
|---|---|---|
| POST | /research | Submit a new company research request |
| GET | /research/{job_id}/report | Poll for completed research report |
| POST | /generate-pdf | Generate PDF from report content |
Required Environment Variables: TAVILY_API_KEY, GEMINI_API_KEY, OPENAI_API_KEY, VITE_GOOGLE_MAPS_API_KEY (frontend). Optional: MONGODB_URI (enable persistent storage).
A complete research cycle consumes approximately 30 Tavily API credits.
Unconfirmed Items
- README footer claims MIT License, but GitHub page and LICENSE file show Apache-2.0; currently recorded per the LICENSE file
- The "GPT-5.1" model name referenced in code has not been publicly confirmed by OpenAI; may be an anticipated name or custom configuration
- Repository contains
langgraph.jsonandlanggraph_entry.pyindicating LangGraph Platform support, but specific deployment steps are not detailed in README