NYC Property Intel

NYC Property Intel

MCP server that gives Claude AI access to 20+ NYC public record datasets for real estate due diligence. Ask Claude about any NYC property in plain English — violations, liens, sales history, ownership, permits, rent stabilization, zoning, fire history, crime data, and more.

This is a due diligence tool, not an appraisal tool. It surfaces public record data only; it does not estimate property values.

Quickstart — Hosted (no local setup)

Sign up at nycpropertyintel.com to get a free trial token (10 queries/day for 30 days, including up to 5 full due-diligence reports). Then:

Claude Code — run once in your terminal:

claude mcp add --transport http nyc-property-intel \
  "https://nyc-property-intel-production.up.railway.app/mcp" \
  --header "Authorization: Bearer YOUR_TOKEN" \
  --scope user

Claude Desktop — add to claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "nyc-property-intel": {
      "type": "http",
      "url": "https://nyc-property-intel-production.up.railway.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Then ask Claude:

"Look up 350 5th Ave, Manhattan"
"What violations does 123 Atlantic Ave, Brooklyn have?"
"Full due diligence on 123 Atlantic Ave, Brooklyn"

Self-Hosting

Prerequisites

• Python 3.12+
• uv package manager
• PostgreSQL 16+
• nycdb CLI (pip install nycdb)

1. Clone and install

git clone https://github.com/ccedacero/nyc-property-intel.git
cd nyc-property-intel
uv sync

2. Configure environment

cp .env.example .env
# Edit .env:
#   DATABASE_URL=postgresql://nycdb:nycdb@localhost:5432/nycdb
#   NYC_GEOCLIENT_SUBSCRIPTION_KEY=your_key  (optional, improves address resolution)
#   SOCRATA_APP_TOKEN=your_token             (optional, higher rate limits for 311/FDNY/NYPD)

3. Set up the database

Option A: Restore from dump (~10 min)

createuser -s nycdb 2>/dev/null; createdb -O nycdb nycdb 2>/dev/null
pg_restore -U nycdb -d nycdb --no-owner --jobs=4 data/nycdb.dump

Option B: Load from source (~2.5 hours)

createuser -s nycdb 2>/dev/null; createdb -O nycdb nycdb 2>/dev/null
chmod +x scripts/seed_nycdb.sh
./scripts/seed_nycdb.sh              # all phases
# or incrementally:
./scripts/seed_nycdb.sh --phase A    # core data (~30 min)
./scripts/seed_nycdb.sh --phase B    # sales + DOB (~45 min)
./scripts/seed_nycdb.sh --phase C    # ACRIS + permits (~90 min)

Then create indexes and materialized views:

psql -U nycdb -d nycdb -f scripts/create_indexes.sql
psql -U nycdb -d nycdb -f scripts/create_views.sql

4. Add to Claude Desktop

{
  "mcpServers": {
    "nyc-property-intel": {
      "command": "uv",
      "args": ["run", "nyc-property-intel"],
      "cwd": "/absolute/path/to/nyc-property-intel",
      "env": {
        "DATABASE_URL": "postgresql://nycdb:nycdb@localhost:5432/nycdb"
      }
    }
  }
}

5. Add to Claude Code

The .mcp.json in the project root auto-registers when you open this directory. Or add manually to ~/.claude.json:

{
  "mcpServers": {
    "nyc-property-intel": {
      "command": "uv",
      "args": ["run", "nyc-property-intel"],
      "cwd": "/absolute/path/to/nyc-property-intel"
    }
  }
}

18 Tools

Data Sources

Core data (~19 million rows) is loaded from nycdb. All datasets are loaded into PostgreSQL. Socrata API used as fallback only.

Architecture

src/nyc_property_intel/
  app.py              # FastMCP instance + system prompt
  server.py           # Entry point: lifespan, auth middleware, tool registration
  config.py           # pydantic-settings (reads .env / environment variables)
  db.py               # asyncpg connection pool + query helpers
  auth.py             # Token validation, rate limiting, usage logging
  analytics.py        # PostHog event capture (fire-and-forget)
  geoclient.py        # NYC GeoClient API + PAD fallback for address resolution
  socrata.py          # Socrata Open Data API client — fallback for FDNY, 311, NYPD, evictions
  loops_webhook.py    # Loops.so webhook → auto-provision trial tokens on signup
  tools/
    lookup.py         # lookup_property
    issues.py         # get_property_issues
    history.py        # get_property_history
    hpd_complaints.py # get_hpd_complaints
    hpd_litigations.py# get_hpd_litigations
    hpd_registration.py# get_hpd_registration
    permits.py        # get_building_permits
    liens.py          # get_liens_and_encumbrances
    tax.py            # get_tax_info
    rentstab.py       # get_rent_stabilization
    comps.py          # search_comps
    neighborhood.py   # search_neighborhood_stats
    fdny.py           # get_fdny_fire_incidents
    complaints_311.py # get_311_complaints
    evictions.py      # get_evictions
    dob_complaints.py # get_dob_complaints
    nypd_crime.py     # get_nypd_crime
    analysis.py       # analyze_property
scripts/
  seed_nycdb.sh         # Downloads and loads all nycdb datasets
  create_indexes.sql    # Performance indexes on critical columns
  create_views.sql      # Materialized views for fast property lookup
  manage_tokens.py      # CLI for provisioning and managing customer tokens

Development

uv run pytest tests/test_utils.py -q          # unit tests (no DB needed)
uv run pytest tests/ -m integration -q        # integration tests (needs live DB)
uv run ruff check src/                         # lint
uv run nyc-property-intel                      # run server locally (stdio)

Fair Housing

NYC Property Intel provides building and property data from public city records only. It does not provide demographic data, tenant screening, or any analysis based on protected characteristics. See nycpropertyintel.com/#fair-housing for the full policy.

License

MIT