Property research with Thirdwatch MCP: listings and sample statistics

Property listings can answer practical questions about asking prices and disclosed features. They cannot by themselves tell you what properties sold for or whether an investment is suitable. Thirdwatch offers two different tools: Indian listing search and sample-based property market analysis.

Build an Indian listing shortlist

search_properties accepts an Indian city, rental/buying/PG mode and listing filters. For example:

{
  "city": "Bangalore",
  "search_mode": "rent",
  "localities": ["Koramangala"],
  "bhk": [2],
  "max_price": 60000,
  "max_results": 10
}

Rent filters use INR per month; buying filters use total asking price. Ask for the price unit in the output. This listing tool is India only and requests data from NoBroker, MagicBricks, CommonFloor and 99acres. Its configured charge is 8 credits for the source fan-out; confirm current pricing with the tool preview before running it.

Try this instruction:

Compare the returned listings by asking rent, locality, furnishing and disclosed area. Preserve listing URLs and source names. Keep missing deposit and area values blank. Separate carpet area from super area and flag source failures.

These are example inputs, not current properties for rent. Contact the listing owner or agent to verify availability, price, fees and property details before relying on a shortlist. Repeated or stale advertisements may remain in the returned sample.

Ask a different tool for market statistics

analyze_property_market returns an aggregate sample, rather than the individual listing table. Its supported geographies are India and the UK. Indian analysis uses NoBroker, MagicBricks and 99acres; UK analysis uses Rightmove.

{
  "city": "London",
  "country": "UK",
  "listing_type": "rent",
  "bedrooms": [2],
  "max_results": 20,
  "compute_yield": false
}

The configured base charges are 6 credits for India and 2 for the UK. Coverage and prices can change; inspect the returned metadata. Ask the assistant to show sample size, median, p25 and p75, currency and missing fields. A small or uneven source sample should not be presented as a complete market index.

Understand the rental-yield option

Setting compute_yield to true fetches the opposite listing mode as well and doubles the configured source charge. The computed gross yield is annualized median asking rent divided by median asking sale price. It compares two samples, not necessarily matched properties.

That calculation omits vacancy, financing, transaction fees, taxes, maintenance and property-specific differences. It is a screening statistic, not a valuation, return forecast or investment recommendation. Ask for both sample sizes and their coverage before interpreting it.

Keep an audit trail

Record the city, locality, mode, bedroom filters, request date, source names and warnings. If you need to repeat a search, use the same input and compare the source coverage first; a changed median may come from a changed sample.

Explore the property workflow, connect your assistant and review pricing. Source-specific Apify tutorials remain available for standalone Actor exports; their inputs and billing differ from the Thirdwatch MCP tools.