Build Your Underwriting Agent

You need Node.js 18 or higher. Check your version by opening Terminal (Mac) or Command Prompt (Windows) and running:

node --version

If you don't have it or it's below 18, install it from nodejs.org/download ↗. Download the LTS version. Run the installer. That's it.

Now create your project:

mkdir underwriting-agent
cd underwriting-agent
npm init -y
npm install @anthropic-ai/sdk pdf-parse csv-parse dotenv exceljs

What each package does:

Go to console.anthropic.com ↗. Create an account if you don't have one.

Navigate to Settings → API Keys ↗. Click Create Key. Name it underwriting-agent. Copy the key.

In your project folder, create a file called .env:

ANTHROPIC_API_KEY=sk-ant-your-key-here

Add credits to your account:

Create a file called criteria.md in your project folder. This is the most important file. It's YOUR investment thesis in plain English. Claude will evaluate every property against this document.

Start with this template. Replace every bracketed value with your actual numbers:

# My Underwriting Criteria

## Deal Killers (auto-reject if any are true)
- Location outside: [City/State or list of zip codes]
- Property type not: [SFR / Multi 2-4 / Multi 5+ / Commercial]
- Asking price above: $[MAX PRICE]
- Environmental issues (Phase I required)
- Structural issues requiring > $[AMOUNT] to repair
- [Add your own deal killers]

## Scoring Criteria (0-10 each)
- Cap Rate: Target [X]%+
  (10 = above target, 5 = at target, 0 = below [Y]%)
- Cash-on-Cash Return: Target [X]%+
- Occupancy Rate: Target [X]%+
  (10 = 95%+, 5 = 85%, 0 = below 70%)
- Rent-to-Market Ratio: How far below market?
  (10 = 20%+ below market, 5 = at market, 0 = above market)
- Neighborhood: A/B/C/D
  (10 = A, 7 = B, 4 = C, 0 = D)
- Deferred Maintenance: Relative to purchase price
  (10 = cosmetic only, 5 = moderate, 0 = major systems)
- Value-Add Potential: Can rents increase? Units added?
  (10 = multiple value-add plays, 0 = fully optimized)
- Financing: Seller financing? Assumable loan?
  (10 = seller financing available, 5 = conventional, 0 = hard money only)

## Scoring Weights
- Cap Rate: 20%
- Cash-on-Cash: 15%
- Occupancy: 15%
- Rent Upside: 15%
- Neighborhood: 10%
- Maintenance: 10%
- Value-Add: 10%
- Financing: 5%

## Red Flags (flag but don't auto-reject)
- Property taxes increased > 20% in last 2 years
- Insurance claims history
- Vacancy above 30%
- Negative cash flow at current rents
- Seller owned less than 2 years
- Flood zone or environmental concerns nearby

## Output I Want
For each property:
1. PASS / FAIL / REVIEW
2. Score 0-100 (weighted)
3. Top 3 strengths
4. Top 3 risks
5. Red flags found
6. Estimated rehab cost range
7. Projected Year 1 NOI
8. BUY / PASS / NEEDS_MORE_INFO

Tips for writing great criteria:

Create a folder called properties in your project. Drop your property reports into it.

underwriting-agent/
  .env
  criteria.md
  analyze.js          (will be created by Claude Code)
  properties/
    123-main-st.pdf   (offering memorandum)
    456-oak-ave.csv   (rent roll export)
    789-elm.xlsx      (broker's pro forma)
    deal-pipeline.csv (multiple properties in one file)

Supported file types:

How to get property reports:

Scroll down to unlock the full prompt below. Then follow these three steps:

What the generated script does:

Running it:

# Analyze all properties in the folder
node analyze.js ./properties/

# Output will look like:
# ✓ Analyzed 47 properties in 12.3 seconds
# Cost: $0.72
#
# RECOMMENDED (8):
#  1. 123 Main St — Score: 91 — BUY
#  2. 456 Oak Ave — Score: 87 — BUY
#  ...
#
# REVIEW (12):
#  9. 789 Elm St — Score: 68 — NEEDS_MORE_INFO
#  ...
#
# PASS (27):
#  21. 101 Pine Rd — Score: 34 — PASS
#  ...
#
# Results saved to results.json and summary.csv

Open summary.csv in Excel or Google Sheets to see your ranked pipeline with all scores, flags, and recommendations in a sortable table.

After your first batch, review the results critically:

Build your weekly analysis routine:

Advanced: automate the pipeline