mcp-property-server

arti-cat/mcp-property-server

3.2

If you are the rightful owner of mcp-property-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

The Property MCP Server is a FastMCP server designed for querying property listings, offering a robust solution for real estate data management.

Tools
3
Resources
0
Prompts
0

Property MCP Server

A FastMCP server for querying property listings with 475 properties.

✨ Now with ChatGPT UI Widget! Interactive property cards with favorites, sorting, and dark mode support.

🎯 NEW: Lead Capture & CRM! Capture leads, match clients to properties, schedule viewings, and manage your sales pipeline. See

🎨 Reusable Widget Templates! Complete patterns and templates for building your own ChatGPT widgets. See

Quick Start

# Install dependencies
pip install -r requirements.txt

# Run with STDIO (for Claude Desktop)
python3 server.py

# Run with HTTP (for remote access)
python3 server.py --http

Tools

Property Search Tools

get_schema()

Returns the data schema for property listings.

query_listings(...)

Search and filter properties.

Parameters:

  • postcode - Filter by postcode (e.g., "DY4 7LG")
  • property_type - Filter by type (e.g., "Flat")
  • max_price - Maximum price
  • min_bedrooms - Minimum bedrooms
  • has_garden - Must have garden
  • has_parking - Must have parking
  • limit - Max results (default: 5)
calculate_average_price(...)

Calculate average price for matching properties.

Parameters:

  • postcode - Filter by postcode
  • property_type - Filter by type

Lead Capture & CRM Tools

capture_lead(...)

Capture new buyer or seller leads from conversations.

Parameters:

  • full_name, email, mobile, role (required)
  • stage - Lead stage (hot/warm/cold/instructed/completed)
  • budget_max, min_bedrooms - For buyers
  • selling_property_id, asking_price - For sellers
match_client(...)

Find properties matching a buyer's preferences.

Parameters:

  • client_id (required) - Buyer's client ID
  • limit - Max results (default: 10)
schedule_viewing(...)

Book property viewings with conflict detection.

Parameters:

  • property_id, buyer_client_id, datetime_iso (required)
  • notes - Optional viewing notes
view_leads(...)

View and filter client pipeline.

Parameters:

  • role - Filter by buyer/seller
  • stage - Filter by lead stage
  • limit - Max results (default: 20)

Testing

# Install test dependencies
pip install pytest pytest-asyncio

# Run tests
python3 -m pytest test_server.py -v

Transports

STDIO (default) - For Claude Desktop, Cursor

python3 server.py

HTTP - For remote access

python3 server.py --http
# Server at http://127.0.0.1:8000/mcp

Project Structure

├── server.py           # Main FastMCP server
├── tools.py            # Tool implementations
├── data_loader.py      # Data loading
├── test_server.py      # Pytest tests
├── requirements.txt    # Dependencies
├── web/                # React UI widget (NEW!)
│   ├── src/            # React components
│   ├── dist/           # Built bundle
│   └── test.html       # Local testing
├── docs/               # Documentation
│   ├── WIDGET_IMPLEMENTATION_PLAN.md
│   ├── WIDGET_DEPLOYMENT.md
│   └── WIDGET_SUMMARY.md
└── data/
    └── listings.jsonl  # Property data (475 listings)

Claude Desktop Setup

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "property-server": {
      "command": "python3",
      "args": ["/absolute/path/to/server.py"]
    }
  }
}

ChatGPT Setup

Requirements

  • ChatGPT Pro, Team, Enterprise, or Edu account
  • Developer Mode enabled in ChatGPT settings

Setup Steps

  1. Start the server with HTTP transport:

    python3 server.py --http

  2. Expose with ngrok:

    ngrok http 8000

    Note your public URL (e.g., https://abc123.ngrok-free.dev)

  3. Enable Developer Mode in ChatGPT:

    • Go to Settings → Connectors
    • Under Advanced, toggle Developer Mode to enabled

  4. Create Connector:

    • In Settings → Connectors, click Create
    • Name: Property Server
    • Server URL: https://your-ngrok-url.ngrok-free.dev/mcp/
    • Check I trust this provider
    • Click Create

  5. Use in Chat:

    • Start a new chat
    • Click + → More → Developer Mode
    • Enable your Property Server connector
    • Ask questions like:
      • "Show me properties in DY4 7LG under £100,000"
      • "Find flats with parking"
      • "What's the average price for 2-bedroom properties?"

Features

  • ✅ All 3 tools have readOnlyHint annotations (no confirmation prompts)
  • ✅ Natural language queries
  • ✅ 475 property listings
  • ✅ Filter by postcode, type, price, bedrooms, garden, parking

UI Widget

The server now includes an interactive React widget for ChatGPT:

Features:

  • 🏠 Property cards with images
  • ❤️ Favorite properties (persisted)
  • 🔄 Sort by price or bedrooms
  • 🌓 Dark mode support
  • 📱 Responsive design

Quick Start:

# Build widget
cd web && npm install && npm run build

# Test locally
open test.html

# Deploy to ChatGPT
# See docs/WIDGET_DEPLOYMENT.md

Documentation:

Widget Templates

This project includes complete reusable templates for building ChatGPT Apps SDK widgets:

  • - Overview & quick start
  • - Complete implementation guide
  • - Quick reference
  • - System architecture
  • - Project generator

Create a new widget project:

cd docs
./create-widget-project.sh my-widget-name

Production Deployment

Deploy to persistent hosting with automatic HTTPS:

# Quick deploy to Fly.io
./deploy.sh

# Manual deployment
fly launch --no-deploy
fly deploy

Supported Platforms:

  • ✅ Fly.io (recommended) - Free tier available
  • ✅ Render - Auto-deploy from Git
  • ✅ Railway - Simple pricing
  • ✅ Google Cloud Run - Scale-to-zero

See for complete guide.

Resources