Agentic Venture Studio — Layer 1: Opportunity Engine

A multi-agent system that continuously scans markets, generates investment theses, and validates ventures opportunities.

Architecture

┌─────────────────────────────────────────────────┐
│              ORCHESTRATOR (DAG Scheduler)        │
│  Manages workflow: scan → generate → validate    │
│  Implements signal bus + CoS governance checks   │
├─────────────┬───────────────┬───────────────────┤
│   MARKET    │    THESIS     │    VALIDATION     │
│   SCANNER   │   GENERATOR   │      AGENT        │
│             │               │                   │
│ • Web scan  │ • Synthesize  │ • Stress-test     │
│ • Signals   │ • Thesis doc  │ • Score 0-1       │
│ • Classify  │ • Abundance   │ • Kill/continue   │
│ • Rank      │   angle       │ • Evidence pack   │
└─────────────┴───────────────┴───────────────────┘

Agents

Agent	Role	Tools	Output
Market Scanner	Scans patent filings, funding announcements, regulatory changes, research papers, job postings to identify emerging opportunity spaces	web_search, extract_signals, classify_signal	Ranked market signals
Thesis Generator	Synthesizes raw signals into structured investment theses with abundance-economy framing	generate_thesis, score_abundance_fit	Thesis documents
Validation Agent	Stress-tests theses against market data, competitive landscape, regulatory risk, and technical feasibility	validate_market, assess_competition, score_feasibility	Validation verdicts

Setup

# 1. Clone and enter directory
cd opportunity_engine

# 2. Create virtual environment
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

# 3. Install dependencies
pip install -r requirements.txt

# 4. Configure API key
cp .env.example .env
# Edit .env and add your ANTHROPIC_API_KEY

# 5. Run the full pipeline
python main.py

# 6. Run individual agents
python main.py --agent scanner --query "AI robotics 2026"
python main.py --agent thesis --signal-file output/signals_latest.json
python main.py --agent validator --thesis-file output/theses_latest.json

# 7. Run with a specific abundance idea
python main.py --idea "autonomous micro-factory network"

Configuration

Edit config.py to adjust:

• Model selection (default: claude-sonnet-4-6)
• Max tokens per agent call
• Number of signals to scan
• Validation thresholds
• Abundance Threshold (IER floor)

Output

All outputs are saved to output/ as timestamped JSON files:

• signals_YYYYMMDD_HHMMSS.json — Raw market signals
• theses_YYYYMMDD_HHMMSS.json — Generated theses
• validations_YYYYMMDD_HHMMSS.json — Validation verdicts
• pipeline_YYYYMMDD_HHMMSS.json — Full pipeline run with audit trail

Stack

• Anthropic Python SDK — Claude Sonnet 4.6 API with tool use
• Pydantic — Structured data models & validation
• asyncio — Concurrent agent execution
• Rich — Terminal UI for pipeline monitoring