Startup Sim Agent

"Turn an idea into a stress-tested startup plan with autonomous agents."

A hackathon-ready monorepo that takes a startup idea and produces a comprehensive analysis using multi-agent orchestration with CrewAI, market intelligence, knowledge graphs, and financial modeling.

Features

• 🤖 10 Specialized AI Agents orchestrated with CrewAI
• 🔍 Live Market Research with Tavily API + citations
• 🕸️ Knowledge Graph in Neo4j for competitor mapping
• 💰 Financial Modeling with unit economics
• 🎯 Investor Debate (Bull vs Skeptic + synthesis)
• 📊 GO/NO-GO Scorecard with next experiments
• 🔄 Real-time Streaming progress updates via SSE
• 📦 Production-Ready with AWS (DynamoDB, S3, Lambda)

Tech Stack

• Frontend: React 18 + Vite + Tailwind CSS
• Backend: FastAPI + Python 3.11
• Agents: CrewAI + OpenAI GPT-4
• Search: Tavily API
• Database: Neo4j (graph) + DynamoDB (state) + SQLite (local)
• Storage: S3 (prod) + filesystem (local)
• Optional: Yutori, Senso, Modulate, Numeric (with adapters)
• Deploy: Render + AWS CDK

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Docker & Docker Compose
• Make

1. Clone and Setup

git clone <repo>
cd startup-sim-agent
cp .env.example .env
# Edit .env with your API keys

2. Install Dependencies

make install

3. Start Local Development

make dev

This starts:

• Frontend: http://localhost:5173
• API: http://localhost:8000
• Neo4j: http://localhost:7474

4. Run a Simulation

1. Open http://localhost:5173
2. Enter a startup idea: "AI-powered meal planning for busy parents"
3. Click "Run Simulation"
4. Watch live progress as agents work
5. Download the final report

Environment Variables

See .env.example for all required keys:

Required:

• OPENAI_API_KEY - GPT-4 access
• TAVILY_API_KEY - Market research

Optional (with stubs):

• YUTORI_API_KEY - Deep web extraction
• SENSO_API_KEY - RAG knowledge base
• MODULATE_API_KEY - Content moderation
• NUMERIC_API_KEY - Financial templates
• NEO4J_URI, NEO4J_USER, NEO4J_PASSWORD
• AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION

API Endpoints

• POST /api/runs - Start new simulation
• GET /api/runs/{run_id}/events - SSE stream
• GET /api/runs/{run_id} - Get dossier JSON
• GET /api/runs/{run_id}/artifact/report.md - Download markdown
• GET /api/runs/{run_id}/artifact/report.pdf - Download PDF
• POST /api/runs/{run_id}/ask - RAG Q&A (if Senso enabled)

Agent Workflow

1. ClarifierAgent - Structure the raw idea
2. MarketResearchAgent - Find competitors + citations
3. PositioningAgent - Define ICP + differentiation
4. MVPPlannerAgent - 4-week roadmap
5. LandingCopyAgent - Marketing copy + pricing
6. BullInvestorAgent - Argue upside
7. SkepticInvestorAgent - Attack weaknesses
8. ModeratorAgent - Synthesize debate
9. FinanceAgent - Unit economics
10. FinalizerAgent - GO/NO-GO + experiments

Development Commands

make install      # Install all dependencies
make dev          # Start dev environment
make test         # Run tests
make lint         # Lint code
make format       # Format code
make clean        # Clean artifacts

Deployment

Render (Frontend + API)

# Connect your repo to Render
# It will auto-detect render.yaml

AWS (Storage + Lambda)

cd infra/aws
npm install
npx cdk deploy

Demo Script (for Judges)

1. Show the problem: "I have a startup idea but don't know if it's viable"
2. Enter idea: "AI-powered legal document analyzer for small businesses"
3. Live progress: Point out each agent step streaming in real-time
4. Market graph: Show Neo4j browser with competitor relationships
5. Debate: Highlight bull vs skeptic arguments
6. Scorecard: Show GO/NO-GO recommendation with reasoning
7. Download: Export full report as PDF
8. Previous runs: Show run history and persistence

Architecture Highlights

• Stateful workflow: Single StartupDossier JSON updated by each agent
• Streaming: SSE for real-time progress updates
• Citations: Every market claim links to source URLs
• Graph DB: Neo4j stores competitive landscape relationships
• Modular integrations: Stub adapters for optional services
• Production-ready: AWS infrastructure with DynamoDB + S3

License

MIT