build-with-ag2

Due Diligence Agent System: AG2 + TinyFish

• Created by John Marshall
• Last revision: 03/11/2026

A multi-agent due diligence pipeline that automatically researches a company from a single URL. It uses AG2 to orchestrate specialist agents in parallel threads, each powered by TinyFish for deep web scraping.

Overview

Given a company URL, the system runs a 4-stage pipeline:

1. Seed Crawler — Scrapes the company website to build an initial profile (name, description, team pages, press pages, job URLs, etc.)
2. Parallel Specialists — Spawns 6 specialist agents concurrently, each using TinyFish to deep-scrape relevant sources:
   • Founders & Team — LinkedIn, about/team pages
   • Investors & Funding — Crunchbase, investor pages
   • Press Coverage — Google News, company press pages
   • Financials — Yahoo Finance, Crunchbase
   • Technology Stack — BuiltWith, GitHub, job postings, engineering blogs
   • Social Signals — LinkedIn, Twitter/X, GitHub
3. Validator — Cross-checks all collected data for contradictions, gaps, and low-confidence fields
4. Synthesis — Produces a structured markdown due diligence report

After the pipeline completes, an interactive Q&A mode lets you ask follow-up questions grounded in the collected data.

AG2 Features

• ConversableAgent — Each specialist is an AssistantAgent with a focused system prompt
• TinyFishTool API — AG2’s built-in TinyFish integration registered as a callable tool for agents

Tags

TAGS: due-diligence, multi-agent, web-scraping, tinyfish, parallel-agents, research-assistant, company-research, automation

Installation

Prerequisites

• Python 3.10+
• An OpenAI API key
• A TinyFish API key

Setup

1. Clone and navigate to the folder:

   git clone https://github.com/ag2ai/build-with-ag2.git
   cd build-with-ag2/due-diligence-with-tinyfish
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Set environment variables:

   export OPENAI_API_KEY=your-openai-key
   export TINYFISH_API_KEY=your-tinyfish-key
   

Usage

Run the full pipeline

python main.py --url https://example.com

This will:

• Run all 4 stages of the pipeline
• Save structured JSON outputs and a final report to a timestamped directory (e.g., due_diligence_acme_20260311_120000/)
• Enter interactive Q&A mode

Q&A on an existing report

python main.py --report-path ./due_diligence_acme_20260311_120000/

Skip the pipeline and jump straight into Q&A over a previously generated report.

Output Structure

due_diligence_acme_20260311_120000/
├── company_profile.json      # Seed crawl results
├── founders_team/
│   ├── founders.json
│   ├── executives.json
│   └── headcount.json
├── investors.json
├── press/
│   ├── articles.json
│   └── sentiment.json
├── financials.json
├── tech_stack.json
├── social.json
├── validation_notes.json
├── report.md                 # Final synthesized report
└── references.md             # Index of all output files

Contact

For more information or any questions, please refer to the documentation or reach out to us!

• View Documentation at: https://docs.ag2.ai/latest/
• Find AG2 on GitHub: https://github.com/ag2ai/ag2
• Join us on Discord: https://discord.gg/pAbnFJrkgZ

License

This project is licensed under the Apache License 2.0. See the LICENSE for details.

This site is open source. Improve this page.