# PropFlow — Real Estate Business Automation
# Claude Code Project Memory — Keep updated after each session

## Project Overview
This is a real estate investment automation system covering 4 strategies:
wholesaling, fix-and-flip, buy-and-hold rentals, and multifamily.
All automation is built with Claude Code + n8n + Supabase.

## Business Context
- Primary Strategy: Wholesaling (low-capital start)
- Secondary: BRRRR (Buy, Rehab, Rent, Refinance, Repeat) rentals for long-term wealth
- Target Market: [YOUR MARKET ZIP CODES]
- Business Entity: [YOUR LLC NAME] — Single Member LLC (Limited Liability Company)
- Bank Account: Dedicated business checking — never mix personal

## Tech Stack
- Automation: n8n (self-hosted on Render/Railway)
- Database: Supabase (PostgreSQL) — leads, deals, contacts, financials
- AI Layer: Claude API (Anthropic SDK) — analysis, document generation
- CRM (Customer Relationship Management): REsimpli OR Left Main REI — deal pipeline
- Skip Tracing: BatchLeads API / REISkip API
- SMS: Twilio — automated sequences
- Email: SendGrid — drip campaigns
- Documents: Docupilot + Dotloop — contracts and e-signing
- Accounting: QuickBooks Online + Stessa (free for rentals)
- Property Mgmt: TurboTenant / Buildium

## Project Structure
propflow/
├── CLAUDE.md                  ← This file (project memory)
├── agents/
│   ├── supervisor.py          ← Main orchestrator
│   ├── lead_agent.py          ← Lead gen + skip tracing
│   ├── analysis_agent.py      ← Deal analysis + underwriting
│   ├── docs_agent.py          ← Contract + document generation
│   ├── marketing_agent.py     ← Outreach automation
│   └── compliance_agent.py    ← Tax + legal monitoring
├── workflows/
│   ├── wholesaling/           ← n8n workflows for wholesaling
│   ├── fix_flip/              ← n8n workflows for flips
│   ├── rentals/               ← BRRRR tracking workflows
│   └── multifamily/           ← Syndication / underwriting
├── calculators/
│   ├── mao_calculator.py      ← MAO (Maximum Allowable Offer) = ARV (After Repair Value) × 0.70 − Repairs
│   ├── brrrr_analyzer.py      ← BRRRR cash-out refi logic
│   ├── cap_rate.py            ← NOI (Net Operating Income) / Property Value
│   └── dscr_checker.py        ← Rental DSCR (Debt Service Coverage Ratio) calculations
├── templates/
│   ├── loi_template.md        ← Letter of Intent template
│   ├── assignment_contract.md ← Wholesale assignment
│   └── purchase_agreement.md  ← Standard purchase template
└── mcp/
    └── mcp_config.json        ← MCP tool connections

## Key Commands
python agents/supervisor.py --task "analyze 123 Main St, Dallas TX"
python agents/lead_agent.py --market "75201" --type absentee_owners
python calculators/mao_calculator.py --arv 250000 --repairs 40000
python workflows/wholesaling/run_pipeline.py

## Coding Conventions
- Use async/await for all API calls
- Log every action to Supabase `workflow_logs` table
- NEVER commit API keys — use .env files
- Always validate property data before running analysis
- Format output as structured JSON for CRM push
- Follow Explore → Plan → Code → Commit workflow

{
  "mcpServers": {
    "supabase": {
      // Leads DB, deal pipeline, workflow logs
      "command": "npx @modelcontextprotocol/server-postgres",
      "env": { "DATABASE_URL": "$SUPABASE_URL" }
    },
    "github": {
      // Version control for workflows and templates
      "command": "npx @modelcontextprotocol/server-github",
      "env": { "GITHUB_TOKEN": "$GITHUB_PAT" }
    },
    "filesystem": {
      // Access property PDFs, OMs, contracts locally
      "command": "npx @modelcontextprotocol/server-filesystem",
      "args": ["/propflow/documents"]
    },
    "notion": {
      // SOPs, team docs, investor updates
      "command": "npx @modelcontextprotocol/server-notion",
      "env": { "NOTION_TOKEN": "$NOTION_API_KEY" }
    }
  }
}

import anthropic
import asyncio

# Initialize Claude client
client = anthropic.Anthropic()

SYSTEM_PROMPT = """You are PropFlow's Real Estate Analysis Agent.
Given a property address, you will:
1. Pull comparable sales (comps) from available market data
2. Calculate After Repair Value (ARV) with confidence range
3. Estimate repairs using standardized cost database
4. Apply the 70% Rule: MAO = ARV × 0.70 - Repair Costs
5. Build a 3-scenario pro forma (conservative/base/optimistic)
6. Generate a DD checklist with 65 items and risk flags
7. Draft an LOI at the calculated MAO

Strategy context: {strategy}  (wholesale/flip/rental/multifamily)
Market: {market}
Investor buy box: Max repairs ${max_repairs}K, Min spread ${min_spread}K

Output format: Structured JSON + plain-language executive summary."""

async def analyze_deal(address: str, strategy: str, market: str):
    """Full deal analysis with 5 sub-agents running in parallel"""
    message = client.messages.create(
        model="claude-opus-4-5",
        max_tokens=8192,
        system=SYSTEM_PROMPT.format(
            strategy=strategy, market=market,
            max_repairs=60, min_spread=25
        ),
        messages=[{
            "role": "user",
            "content": f"Analyze this property: {address}\nStrategy: {strategy}"
        }]
    )
    return message.content[0].text