AI Software Company

A multi-agent AI pipeline that simulates a full software company — from problem analysis to working code, architecture diagrams, and investor-ready presentations.

Live Demo: https://ai-software-company-gold.vercel.app

How It Works

Enter a problem statement and watch 6 AI agents collaborate in real-time:

Problem Statement
       |
   [CEO Agent] ---- Analyzes problem, classifies deliverable type
       |
   [Business Analyst] ---- Requirements, user stories, acceptance criteria
       |
   [Researcher] ---- Market analysis, tech stack research, web search
       |
   [Architect] ---- System design, API specs, database schema
       |
   [Engineer] ---- Full working code / n8n workflow generation
       |
   [PPT Agent] ---- Investor-ready presentation slides

Each agent's output is cross-reviewed by a peer agent before moving to the next stage. The pipeline supports both auto-pilot (fully autonomous) and manual approval modes.

Features

• Smart Task Routing — Classifies inputs into 5 pipeline routes (full product, workflow, research, presentation, hybrid) with route-specific prompts and guardrails
• Domain Memory — Learns from completed projects and applies domain knowledge to future ones
• OAuth Login — Sign in with GitHub, Google, or email/password
• Live Code Preview — Real-time preview of generated code via E2B sandboxes
• Streaming Agent Chat — Talk to any agent mid-pipeline with streaming responses
• GitHub Push — Push generated code directly to a GitHub repository
• Shareable Links — Share project results via unique URLs
• Build & Test Execution — Sandboxed code execution with build status tracking

Deliverables

After a pipeline run completes, you get:

• Source Code (.zip) — Full project with all files
• Architecture Diagram — Visual system design
• Presentation (.pptx) — Investor/pitch deck
• Project Report (.docx) — Comprehensive documentation
• n8n Workflow (.json) — Importable automation workflow (for workflow/hybrid projects)

Tech Stack

Layer	Technology
Backend	Python 3.11, FastAPI, SQLite / PostgreSQL
Frontend	Next.js 16, TypeScript, Tailwind CSS
Auth	JWT (HS256), GitHub OAuth, Google OAuth
LLM Providers	OpenRouter, OpenAI, Gemini
Real-time	WebSocket with auto-reconnect
Cache	Redis
Workflow RAG	19,800+ n8n workflow templates indexed
Hosting	Vercel (frontend), Railway (backend)

Architecture

• Multi-provider routing — Each agent gets a dedicated API key, with automatic cross-key failover on quota exhaustion (402), rate limits (429), or auth errors (401/403)
• Task routing — Classifies inputs into 5 pipeline routes with route-specific prompts
• Streaming output — Live token-by-token output visible in the dashboard
• Cross-review — Every agent output is peer-reviewed by another agent before approval
• Workflow RAG — Searches 19,800+ n8n workflow templates to find reusable components
• Domain memory — Persists learnings across projects for better context
• Crash-resilient — Debounced state refreshes, capped event streams, connection auto-recovery

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• At least 1 OpenRouter API key

Backend

cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# Create .env with your API keys (see .env.example)
cp .env.example .env
# Edit .env with your keys

python -m uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload

Frontend

cd frontend
npm install

# Point to your backend
cat > .env.local << 'EOF'
NEXT_PUBLIC_API_URL=http://localhost:8000/api
NEXT_PUBLIC_WS_URL=ws://localhost:8000/api
EOF

npm run dev

Open http://localhost:3000 in your browser.

Environment Variables

Variable	Required	Description
OPENROUTER_API_KEY	Yes	Primary OpenRouter key
OPENROUTER_API_KEY_2 to _6	No	Additional keys for per-agent routing
OPENAI_API_KEY	No	Direct OpenAI access for Engineer
GEMINI_API_KEY	No	Direct Gemini access
JWT_SECRET	Prod	Secret for JWT token signing (required in production)
GITHUB_CLIENT_ID	No	GitHub OAuth app client ID
GITHUB_CLIENT_SECRET	No	GitHub OAuth app client secret
GOOGLE_CLIENT_ID	No	Google OAuth app client ID
GOOGLE_CLIENT_SECRET	No	Google OAuth app client secret
FRONTEND_URL	No	Frontend URL for OAuth redirects (default: http://localhost:3000)
BACKEND_URL	No	Backend URL for OAuth callbacks (default: http://localhost:8000)
DATABASE_URL	No	PostgreSQL connection string (uses SQLite if not set)
DATABASE_PATH	No	SQLite database path (default: company.db)
REDIS_URL	No	Redis connection string
SMART_MODEL	No	Default smart model (default: openrouter/free)
FALLBACK_MODEL	No	Default fallback model (default: openrouter/free)
N8N_WEBHOOK_URL	No	n8n webhook for sharing

Multi-Key Setup (Recommended)

For best performance, configure 6 OpenRouter API keys — one per agent:

OPENROUTER_API_KEY=sk-or-v1-key1
OPENROUTER_API_KEY_2=sk-or-v1-key2
OPENROUTER_API_KEY_3=sk-or-v1-key3
OPENROUTER_API_KEY_4=sk-or-v1-key4
OPENROUTER_API_KEY_5=sk-or-v1-key5
OPENROUTER_API_KEY_6=sk-or-v1-key6

API Endpoints

Auth

Method	Endpoint	Description
POST	/api/auth/register	Create account
POST	/api/auth/login	Email/password login
GET	/api/auth/me	Get current user
GET	/api/auth/providers	List enabled OAuth providers
GET	/api/auth/github	Start GitHub OAuth flow
GET	/api/auth/google	Start Google OAuth flow

Projects

Method	Endpoint	Description
POST	/api/projects	Create new project
GET	/api/projects	List all projects
GET	/api/projects/{id}	Get project state
DELETE	/api/projects/{id}	Delete project
POST	/api/projects/{id}/approve/{output_id}	Approve/reject agent output
POST	/api/projects/{id}/revise	Request agent revision

Agent Interaction

Method	Endpoint	Description
POST	/api/projects/{id}/call	Chat with an agent
POST	/api/projects/{id}/call/stream	Streaming agent chat
GET	/api/projects/{id}/conversation/{role}	Get conversation history
POST	/api/classify	Classify input into pipeline route

Downloads & Deliverables

Method	Endpoint	Description
GET	/api/projects/{id}/download/code	Download generated code (.zip)
GET	/api/projects/{id}/download/pptx	Download presentation
GET	/api/projects/{id}/download/docx	Download report
GET	/api/projects/{id}/download/workflow	Download n8n workflow

Code Preview & GitHub

Method	Endpoint	Description
GET	/api/projects/{id}/files	List generated files
GET	/api/projects/{id}/preview	Start live code preview
POST	/api/projects/{id}/push-to-github	Push code to GitHub
POST	/api/settings/github-token	Save GitHub token

Sharing

Method	Endpoint	Description
POST	/api/projects/{id}/share-link	Create shareable link
GET	/api/shared/{token}	View shared project

Real-time

Method	Endpoint	Description
WS	/api/ws/{id}	Real-time project updates

License

MIT