Introduction

Conducting deep business diligence is often a tedious process of jumping between financial reports, news articles, and company websites. The Company Research Agent is an open-source multi-agent tool that automates this workflow, transforming a simple company name into a comprehensive, structured business report. By orchestrating a pipeline of specialized AI agents, it eliminates the manual grind of data gathering and synthesis, allowing analysts and developers to focus on strategic decision-making.

What Is Company Research Agent?

Company Research Agent is a multi-agent AI framework designed to generate detailed company research reports through an automated pipeline. It is built using Python and React, licensed under the Apache License 2.0, and leverages a sophisticated orchestration of specialized research nodes to conduct deep diligence on any organization.

The tool operates as an agentic workflow where separate AI agents are tasked with specific domains of research—such as financial metrics, industry trends, and recent news—before a final editor agent synthesizes the findings into a professional report. This modular architecture ensures that the research is comprehensive and grounded in real-time data retrieved via the Tavily API.

Why Company Research Agent Matters

Traditional company research is fragmented. An analyst typically spends hours searching for a company’s latest funding round, its market position relative to competitors, and recent press releases. This manual process is not only time-consuming but also prone to human error and oversight.

Company Research Agent solves this by providing a unified, agentic approach to business intelligence. Instead of a single LLM attempt to answer a query, the system uses a “divide and conquer” strategy. By delegating tasks to specialized agents (e.g., a Financial Analyst agent and a News Scanner agent), the tool achieves a level of depth and accuracy that general-purpose AI chatbots cannot match.

For developers and business professionals, this tool represents a shift from simple prompt-based AI to agentic workflows. It demonstrates how to combine multiple LLMs—using Gemini 2.5 Flash for high-context synthesis and GPT-5.1 for precise formatting—to create a production-ready business intelligence tool.

Key Features

• Multi-Source Research: Gathers data from a wide array of sources, including official company websites, financial reports, industry analyses, and real-time news articles to ensure a 360-degree view of the target company.
• AI-Powered Content Filtering: Integrates Tavily’s relevance scoring to filter out noise and ensure that only the most pertinent information is passed to the synthesis agents.
• Dual-Model Architecture: Employs a strategic split between models; Gemini 2.5 Flash is used for processing large volumes of research data (high-context synthesis), while GPT-5.1 is utilized for the final report editing and precise formatting.
• Asynchronous Processing: Utilizes a polling-based architecture that allows the system to conduct extensive research in the background while the user tracks progress in real-time via the UI.
• Modern React Frontend: Provides a responsive user interface that allows users to trigger research jobs, monitor the agent’s progress, and download the final generated reports.
• Modular Agent Framework: Built on a pipeline of specialized nodes (Analyzers and Processors) that can be easily extended or modified to add new research dimensions.

How Company Research Agent Compares

While GPT Researcher is a powerful general-purpose autonomous research tool, Company Research Agent is specifically tuned for business intelligence. Its primary differentiator is the use of specialized research nodes—such as the CompanyAnalyzer and FinancialAnalyst—which ensure that business-critical metrics are not overlooked. This specialized focus prevents the “generalization gap” where a general research agent might miss specific financial indicators or industry-specific KPIs.

Compared to a standard LLM interaction, the Company Research Agent removes the hallucination risk associated with static training data. By forcing the agent to retrieve real-time data via Tavily and then synthesize it through a multi-step pipeline, the tool ensures that the final report is grounded in current web evidence rather than probabilistic guesses.

Getting Started: Installation

To run the Company Research Agent locally, you will need API keys for Tavily, Google Gemini, and OpenAI. You should also have Python and Node.js installed on your system.

Quick Setup

The project provides a setup script for rapid deployment:

git clone https://github.com/guy-hartstein/company-research-agent.git
cd company-research-agent
chmod +x setup.sh
./setup.sh

Manual Installation

If you prefer manual control, follow these steps to configure the backend and frontend separately.

Backend Setup

# Create and activate a virtual environment
python -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Frontend Setup

cd ui
npm install

Environment Configuration

You must create .env files for both the backend and the frontend to provide the necessary API keys.

Backend .env

TAVILY_API_KEY=your_tavily_key
GEMINI_API_KEY=your_gemini_key
OPENAI_API_KEY=your_openai_key
MONGODB_URI=your_mongodb_connection_string # Optional

Frontend .env

Create a .env file inside the ui directory:

VITE_API_URL=http://localhost:8000
VITE_GOOGLE_MAPS_API_KEY=your_google_maps_api_key_here

Docker Deployment

The application can be deployed using Docker Compose for a fully containerized environment:

docker-compose up -d

How to Use Company Research Agent

Once the servers are running, the workflow is designed to be intuitive. Start by launching the backend server with python -m application.py and the frontend with npm run dev inside the ui folder.

To generate a report, enter the name or domain of the company you wish to research in the search bar of the web interface. When you click “Run Agent,” the system triggers the research pipeline. The backend will first delegate tasks to the CompanyAnalyzer, IndustryAnalyzer, and FinancialAnalyst agents in parallel.

As the research progresses, the UI will update in real-time to show which agent is currently active and which phase (Research, Curation, Briefing, or Editing) the process is in. Once the process is complete, the system generates a structured markdown report that can be viewed directly in the browser or downloaded for further analysis.

Code Examples

The core of the Company Research Agent is its node-based architecture. Below are examples of how the research nodes are implemented to handle specific business domains.

Research Node Implementation

The CompanyAnalyzer node is responsible for gathering core business information. It uses a structured prompt to guide the LLM to look for specific company details.

# Example logic for a research node
class CompanyAnalyzer:
    def __init__(self, model):
        self.model = model

    def run(self, company_name):
        # The agent uses Tavily to search for core business details
        # and then synthesizes the result into a structured format.
        query = f"What are the core business operations and products of {company_name}?"
        results = tavily.search(query=query, search_depth="advanced")
        return self.model.invoke(results)

Synthesis and Briefing

The Briefing node uses Gemini 2.5 Flash to synthesize large volumes of raw research data into category-specific summaries.

# Example logic for the Briefing node
class BriefingNode:
    def __init__(self, model):
        self.model = model

    def run(self, research_data):
        # Gemini 2.5 Flash handles the high-context synthesis
        prompt = f"Summarize the following research data into a professional business brief: {research_data}"
        return self.model.invoke(prompt)

Real-World Use Cases

Company Research Agent is particularly effective in scenarios where rapid, high-fidelity business intelligence is required without the need for manual searching.

• Sales Prospecting: A sales representative can use the tool to generate a deep-dive report on a prospect before a first call, identifying the company’s current pain points, recent news, and market position to personalize their pitch.
• Competitive Intelligence: A product manager can run the agent against a set of competitors to identify gaps in their product offerings, track their recent pivots, and synthesize a comparison matrix of their strengths and weaknesses.
• Investment Due Diligence: An angel investor or VC analyst can use the tool to quickly vet a startup’s market traction, funding history, and industry trends before deciding to move forward with a deep-dive audit.
• Strategic Planning: A corporate strategist can use the agent to monitor industry shifts and identify emerging competitors by running periodic research reports on key players in a specific sector.

Contributing to Company Research Agent

The project is open-source and welcomes contributions from the community. Since the project follows a modular node-based architecture, the easiest way to contribute is by adding new research nodes or improving the existing prompts for the analyzers.

To contribute, fork the repository and create a feature branch. Submit your changes via a Pull Request. If you encounter bugs or have feature requests, please use the GitHub Issues tab to report them. The project follows the standard GitHub flow for open-source collaboration.

Community and Support

The primary hub for the Company Research Agent community is the GitHub repository. Users can engage in GitHub Discussions to ask questions, share their research reports, and collaborate on new features. The project’s activity level is maintained through active commits and issue tracking on GitHub.

For technical support, refer to the README and the installation guide provided in the repository. The community is focused on developers and AI researchers interested in agentic workflows and business intelligence automation.

Conclusion

The Company Research Agent is a powerful implementation of agentic AI, moving beyond simple chat interfaces to a structured, multi-agent pipeline. By specializing the research process into distinct nodes—Financial, Industry, and News—it provides a level of depth and accuracy that is unattainable with a single LLM prompt.

For analysts and developers who need reliable, real-time business intelligence, this tool is the right choice when you need a structured report rather than a conversational answer. It is a prime example of how to combine the strengths of different LLMs to optimize for both context window and formatting precision.

Star the repo, try the quickstart, and join the community to start automating your business research today.

Resources

• Company Research Agent GitHub Repository
• Tavily AI Search API
• Google Gemini API Documentation