Foundra — AI Startup Operating System

An intelligent, multi-agent AI platform that helps founders validate, research, and build their startup from idea to go-to-market strategy.

---

Overview

Foundra is an agentic AI system designed to act as an operating system for startups. It takes a founder's raw idea and runs it through a structured multi-phase pipeline powered by LLMs (Gemini, Groq, OpenAI, Ollama), LangGraph orchestration, and a suite of specialized AI agents — from problem discovery to go-to-market planning.

Each phase produces a structured JSON report that feeds into the next, enabling end-to-end startup intelligence automation.

---

Architecture

Foundra/
├── app.py                          # CLI entrypoint (phase runner)
├── requirements.txt                # Python dependencies
├── outputs/                        # Phase result JSONs stored here
└── backend/
    ├── app.py                      # FastAPI backend server
    ├── routes/                     # API route handlers
    │   ├── health.py               # Health check endpoint
    │   ├── projects.py             # Project management
    │   ├── phases.py               # Phase execution endpoints
    │   ├── boardroom.py            # Boardroom decision API
    │   └── export.py               # Report export (PDF/PPTX)
    ├── db/                         # Database layer (SQLAlchemy + Supabase)
    └── agentic_engine/
        ├── state.py                # Shared LangGraph state (FoundraState)
        ├── graph.py                # Top-level orchestration graph
        ├── nodes.py                # Graph nodes (worker → executive → board)
        ├── LLMs/                   # LLM provider configurations
        ├── tools/                  # Agent tools (search, scraping, trends)
        ├── bridge/                 # Phase output → API format converters
        ├── executive_agents/       # Executive layer (review & critique)
        ├── boardroom/              # Final decision-making agents
        └── worker_agents/
            ├── phase1_problem_discovery/   # Problem & ICP analysis
            ├── phase2_validation/          # Idea validation
            ├── phase3_market_research/     # TAM/SAM/SOM, competitors
            ├── phase4_business_model/      # Revenue models, pricing
            ├── phase5_product_strategy/    # MVP & roadmap planning
            └── phase6_gtm/                # Go-to-market strategy

---

How It Works

Foundra runs through 6 sequential phases, each powered by dedicated worker agents:

Phase	Name	What it does
1	Problem Discovery	Identifies the core problem, ICP (Ideal Customer Profile), and pain points
2	Validation	Validates demand, checks competition, and assesses feasibility
3	Market Research	Estimates TAM/SAM/SOM, analyzes trends and competitors
4	Business Model	Suggests revenue models, pricing, and monetization strategies
5	Product Strategy	Defines MVP scope, feature roadmap, and tech stack
6	Go-To-Market	Plans launch channels, acquisition funnels, and growth levers

Each phase outputs a structured JSON to outputs/phaseN_result.json, which the next phase consumes. The results are reviewed by executive agents and a final boardroom makes a holistic decision.

---

Tech Stack

Layer	Technology
Orchestration	LangGraph
LLMs	Google Gemini, Groq, OpenAI, Ollama
Web Scraping / Search	Tavily, DuckDuckGo, BeautifulSoup4
Trend Data	PyTrends, Reddit (PRAW)
Backend API	FastAPI + Uvicorn
Database	PostgreSQL (SQLAlchemy + psycopg2) + Supabase
Vector Memory	ChromaDB
Export	ReportLab (PDF), python-pptx (PowerPoint)
Validation	Pydantic

---

Getting Started

1. Clone the repository

git clone https://github.com/your-org/foundra.git
cd foundra

2. Install dependencies

pip install -r requirements.txt

3. Set up environment variables

Create a .env file in the root directory:

GOOGLE_API_KEY=your_google_api_key
GROQ_API_KEY=your_groq_api_key
OPENAI_API_KEY=your_openai_api_key
TAVILY_API_KEY=your_tavily_api_key
SUPABASE_URL=your_supabase_url
SUPABASE_KEY=your_supabase_key
DATABASE_URL=postgresql://user:password@host/dbname

4. Run the backend server

uvicorn backend.app:app --reload --port 8000

5. Run a phase manually (CLI)

Uncomment the relevant phase block in app.py and run:

python app.py

---

API Endpoints

Method	Route	Description
GET	/	Root health check
GET	/health	Service health status
POST	/projects	Create a new startup project
GET	/projects/{id}	Fetch project details
POST	/phases/run	Trigger a phase for a project
GET	/phases/{id}	Get phase result
POST	/boardroom/decide	Trigger boardroom decision
GET	/export/{id}	Export final report (PDF/PPTX)

---

Output Format

Each phase stores results in outputs/phaseN_result.json. Example Phase 1 output:

{
  "startup_name": "EduAI",
  "industry": "EdTech",
  "problem_statement": "Students struggle with personalized exam prep...",
  "icp": { "persona": "College student aged 18-24", ... },
  "pain_points": ["lack of adaptive content", "no real-time feedback"],
  "logs": ["Phase 1 complete"]
}

---

Team

This is a team project built collaboratively by:

Name	GitHub
Rohith	@karnayanarohith
Mani	@mani9441

---

License

This project is currently unlicensed. All rights reserved by the contributors.

---

Built with ❤️ by the Foundra Team