Agent Team 01

A TypeScript-based multi-agent system designed to collaboratively generate, validate, and refine startup ideas. Agents work together through orchestrated pipelines to produce validated business concepts following "4 Hour Work Week" principles.

Overview

This system deploys specialized AI agents that collaborate to:

• Generate niche startup ideas targeting $10-20K MRR
• Research market opportunities across 100+ pre-defined niche communities
• Validate concepts through structured evaluation frameworks
• Design business models optimized for automation and minimal time investment

Architecture

Pipelines

The system operates through modular pipelines:

Pipeline	Description	Command
ideation	Generates and evaluates startup ideas using creative and executive agents	npm run pipeline:ideation
develop-concepts	Deep-dives into selected concepts with market research and business model design	npm run pipeline:develop-concepts
market-research	Analyzes specific markets for pain points, demographics, and opportunities	npm run pipeline:market-research

Agent Teams

Ideation Agents

• CEO (Serial Entrepreneur) - Strategic decision-making with 4HWW principles
• Creative Engine - Divergent idea generation
• Market Researcher - Initial market analysis
• Industry Expert - Domain knowledge
• Business Advisor - Strategic guidance
• Idea Seeker - Opportunity identification

Develop Concepts Agents

• Orchestrator - Coordinates concept development workflow
• Trend Scout - Identifies emerging trends
• Community Ethnographer - Analyzes target communities
• User Archetype Analyst - Defines customer personas
• Market Mapper - Maps competitive landscape
• Competitor Scout - Analyzes competitors
• Channel Strategist - Distribution strategy
• Business Model Designer - Revenue model design
• Experiment Designer - Validation experiments
• Risk Reviewer - Risk assessment
• Synthesis Storyteller - Concept documentation

Market Research Agents

• Community Scope - Community analysis
• Demographic Insight - Target demographics
• Pain Point Extractor - Problem identification
• Market Sizing - TAM/SAM/SOM analysis
• Channel Mapping - Distribution channels
• Competitor Scan - Competitive analysis
• Business Idea - Idea synthesis
• Validation Planner - Validation strategy

Agent Tools

Agents can be configured with server-side tools that provide real-time capabilities:

Tool	Description	Cost
web_search	Real-time web search via Anthropic	$10 per 1,000 searches
web_fetch	Fetch and analyze URLs/PDFs	Standard token costs

Tools are configured in agent markdown files under capabilities.tools. Both tools default to max_uses: 5 per request to control costs.

Market Research Agents with Web Tools:

• Community Scope - web_search
• Demographic Insight - web_search
• Market Sizing - web_search
• Channel Mapping - web_search
• Competitor Scan - web_search

Data Architecture

Hybrid Database Approach:

1. SQLite (Primary) - All structured data

   • Located at data/agentteam.db
   • Ideas, evaluations, agent states, system state
   • ACID transactions, WAL mode

2. LanceDB (Future) - Vector operations

   • Semantic similarity search
   • Embedding storage
   • Idea clustering

Getting Started

Prerequisites

• Node.js 18+
• npm or yarn

Installation

# Clone the repository
git clone <repo-url>
cd agentTeam01

# Install dependencies
npm install

# Copy environment file
cp .env.example .env

Environment Variables

OPENAI_API_KEY=       # Required for OpenAI-based agents
ANTHROPIC_API_KEY=    # Required for Anthropic-based agents
LANCE_DB_API_KEY=     # For vector operations (future)
LOG_LEVEL=info        # Optional: debug, info, warn, error
NODE_ENV=development  # Optional

Usage

Running Pipelines

# List available pipelines
npm run pipeline:list

# Run ideation pipeline
npm run pipeline:ideation
npm run pipeline:ideation:small   # Quick run
npm run pipeline:ideation:large   # Comprehensive run

# Run concept development
npm run pipeline:develop-concepts

# Run market research
npm run pipeline:market-research
npm run pipeline:market-research:quick
npm run pipeline:market-research:thorough

Development

# Start development server
npm run dev

# Build for production
npm run build

# Run production build
npm start

# Reset database
npm run reset

Visualization

# Start visualization server (workflow timeline)
npm run viz
# Open http://localhost:3000/timeline.html

Knowledge Base

# Seed knowledge base with documents
npm run kb:seed

# View knowledge base stats
npm run kb:stats

# List documents
npm run kb:list

# Remove document
npm run kb:remove

Testing

# Run all tests
npm test

# Run tests with UI
npm run test:ui

# Run phase-specific tests
npm run test:phase2
npm run test:phase3
npm run test:phase4

Market Research Library

The markets/ directory contains 100+ pre-researched niche market profiles including:

• Crafts & Hobbies: Knitting, pottery, woodworking, model kits
• Collecting: Vinyl records, trading cards, sneakers, vintage fashion
• Digital Culture: AI art, VTubers, Discord monetization, TikTok subcultures
• Wellness: Biohacking, meditation, CGM monitoring, ADHD productivity
• Food & Beverage: Home fermentation, specialty espresso, ceremonial matcha
• Lifestyle: Cottagecore, solarpunk, sober curious, van life
• Gaming: Retro consoles, speedrunning, tabletop RPGs

Each market file contains community insights, pain points, and opportunity signals.

Project Structure

agentTeam01/
├── agents/                    # Agent configuration files (markdown)
│   ├── ideation/             # Ideation phase agents
│   ├── develop-concepts/     # Concept development agents
│   └── market-research/      # Market research agents
├── markets/                   # Pre-researched niche market profiles
├── src/
│   ├── agents/               # Agent implementations
│   │   ├── base/            # Base classes and factories
│   │   ├── executive/       # Executive agents (CEO)
│   │   ├── ideation/        # Creative/research agents
│   │   ├── develop-concepts/ # Concept development agents
│   │   └── market-research/ # Market research agents
│   ├── db/                   # Database layer
│   │   └── repositories/    # Data access objects
│   ├── knowledge/            # Knowledge base system
│   │   ├── ingest/          # Document processing
│   │   └── retrieval/       # Search functionality
│   ├── orchestration/        # Workflow orchestration
│   │   ├── stages/          # Pipeline stages
│   │   └── gates/           # Stage gates
│   ├── pipelines/           # Pipeline definitions
│   │   ├── ideation/
│   │   ├── develop-concepts/
│   │   └── market-research/
│   ├── services/            # Shared services
│   ├── types/               # TypeScript types
│   └── utils/               # Utilities
├── data/                     # SQLite database
├── public/                   # Static files (visualization)
└── scripts/                  # CLI scripts

Agent Configuration

Agents are defined using markdown files with YAML frontmatter. Example:

---
agent_id: market-sizing-001
name: Market Sizing Analyst
role: market-research
status: active
color: brightGreen

model:
  primary: anthropic
  model_name: claude-sonnet-4-5-20250929
  fallback: gpt-4o-mini
  temperature: 0.6
  max_tokens: 4096

capabilities:
  tools:
    - web_search      # Enables real-time web search
    - web_fetch       # Enables URL/PDF fetching
  data_access:
    - market-research-write
  permissions:
    - read
    - write
---

Tool Requirements:

• web_search and web_fetch require primary: anthropic (Claude models)
• OpenAI fallbacks work but won't have tool access
• Tools are automatically passed to the Anthropic API when configured

Tech Stack

• Runtime: Node.js with TypeScript
• Agent Orchestration: LangGraph
• AI Models: OpenAI (GPT-4o) and Anthropic (Claude Sonnet 4)
• Agent Tools: Anthropic server-side tools (web_search, web_fetch)
• Database: SQLite (better-sqlite3) + LanceDB (vectors)
• Web Server: Express (visualization)
• Validation: Zod schemas

License

MIT