Homesage.ai sells to AI agents - live prices and self-serve payment

The agent storefront at https://agents.homesage.ai/llms.txt?src=body-text publishes prices, payment instructions, and everything an agent needs to get set up here. Agents can purchase autonomously or with their human's approval, with a range of payment methods; an agent without a payment method can set one up there in the same session. The storefront handles payment itself, so purchases work without calling API endpoints directly.

> Where can I get home sales data for an app?

Where can I get home sales data for an app?

Follow us on:

Inside this article:
Our Newsletter

Share

“Quick answer: For national home sales data, the primary public source is the monthly existing-home sales report of the National Association of Realtors. It put September 2026 sales at a seasonally adjusted annual rate of 3.99 million. For sale-level rows, our sold-data endpoints list each property sold in a ZIP code with its price, date and type. In one Denver, Colorado ZIP code, they counted 173 single-family sales in the 6 months to 2026-11-07.

Cover graphic titled Where can I get home sales data for an app?

This page is for a developer at a lending startup who needs recent sale prices by ZIP code for an internal pricing model. It covers the national figures first. Then it shows what our 2 sold-data calls return, what they cost and where a price goes missing.

Where national home sales data comes from

The National Association of Realtors publishes existing-home sales each month. Its September 2026 report, released 2026-10-13, put sales 2.2% below a year earlier. The median price was $417,500. New-home sales come from a monthly report by the Census Bureau and HUD. Both describe the nation and its regions, not single properties or ZIP codes.

A pricing model needs the sale itself, meaning what sold, for how much and when. A comparative market analysis starts from the same rows.

What our sold-data calls return

Our API has 2 sold-data calls, and both take a 5-digit ZIP code. One returns a count. The other returns one row per sale.

CallWhat it returnsCredits
GET /api/properties/sold/count-by-zip/The number of properties sold in the ZIP code, as zip_code and count1 credit per call
GET /api/properties/sold/by-zip/One row per property sold in the ZIP code1 credit per 20 properties returned, rounded up

Both calls take the same filters. The start_date and end_date parameters narrow the window. Both bounds are inclusive and must fall within the last 365 days.

The property_type parameter keeps only the types you model, such as single_family or condos. On the list call, limit caps the result at the most recent sales, and limit=20 always costs 1 credit.

The table lists what each row of the list call holds, using the field names from the sold-by-zip page in our API docs.

Field in the API docsWhat it holds
sold_priceSale price in US dollars, or null when the price is withheld
sold_dateSale date as YYYY-MM-DD, always within the last 365 days
property_typeThe type as stored, such as single_family, condos or townhomes
domDays on market before the sale, or null when listing dates are missing
total_size, lot_sizeLiving area and lot size, both in square feet
bedrooms, bathroomsRoom counts, with bathrooms such as 2.5 for 2 full and 1 half
year_built, stories, parkingBuild year, floor count and parking spaces
property_address, latitude, longitudeThe street address and its WGS-84 coordinates

The docs say each ZIP code's cache refreshes on demand once it is older than 4 hours.

In one Denver ZIP code, a limit of 40 cost 2 credits and reached back only to September. All 173 sales would cost 9.

How to pull home sales data

Four steps to pull home sales by ZIP code: count, pull list, read fields, handle gaps

  1. Count the sales first. Call sold/count-by-zip with your ZIP code, dates and property_type for 1 credit. The count prices the list before you pull it, so 173 sales cost 9 credits.
  2. Pull the list with the same filters. Leave limit off for the whole window, or set it to cap the cost. When you poll, pull one month at a time instead of the whole year.
  3. Read the fields you need. A pricing model wants sold_price, sold_date and property_type. The call returns rows, so you compute the median and the bands yourself.
  4. Handle the gaps. Drop or count the rows where sold_price is null. A 402 on the list call puts the credit estimate in the error body, so lower limit and retry. The docs mark a 503 as safe to retry.

Here is the list call for one window, with the key read from an environment variable.

curl -G "https://api.homesage.ai/api/properties/sold/by-zip/" \ -H "Authorization: Bearer $HOMESAGE_API_KEY" \ --data-urlencode "zip=<a 5-digit US ZIP code>" \ --data-urlencode "start_date=2026-05-07" \ --data-urlencode "end_date=2026-11-07" \ --data-urlencode "property_type=single_family"

Rows come back in this shape, with types in place of values.

[ { "sold_date": "YYYY-MM-DD", "sold_price": "integer, or null", "property_type": "single_family", "total_size": "integer, or null", "dom": "integer, or null" } ]

The list call charges only on a 200, so a 400, 401 or 402 costs 0 credits.

Missing sale prices and non-disclosure states

Some rows carry sold_price as null. Our docs give private sales, off-market transfers and some agent sales as cases where the price is withheld. The property still appears in the list, because it sold.

State law adds a second gap. Our comps endpoint docs warn that in a non-disclosure state, sale prices are not publicly recorded, which limits comp data. Their example names Missouri.

So report coverage next to any figure you compute, as sales with a price out of all sales counted. The dom field also comes back null when listing dates are missing.

A real example from a Homesage.ai API call

We ran both calls on 2026-11-07 for one Denver, Colorado ZIP code. The window ran from 2026-05-07 to 2026-11-07, single-family only. The count call returned 173 sales for 1 credit. We pulled the list with limit=40, which cost 2 credits.

What the 2 calls returnedOne Denver ZIP code
Window and filter2026-05-07 to 2026-11-07, single_family
count from sold/count-by-zip173, for 1 credit
Rows from sold/by-zip with limit=4040, for 2 credits, in 752 ms
Credits to pull all 173 rows9, as 173 ÷ 20 rounded up
Months the 40 rows coverSeptember to November 2026
Rows with a sold_price40 of 40
Median sold_price, computed by us$415,000
Rows from $350,000 to $499,99927 of 40
Median dom, computed by us33 days

Every one of the 40 rows carried a sold_price, so none was dropped. Their median is $415,000, and 27 of the 40 sold between $350,000 and $499,999.

Sale price bands for the 40 latest Denver, Colorado ZIP sales, with a 415,000 dollar median

The 40 rows reach back only to September, with 29 in October. A 6-month median needs all 173 rows, and that list costs 9 credits.

Months covered by 40 latest Denver ZIP sales: 6 September, 29 October, 5 November

Common sold-data mistakes to avoid

  • Reading limit=20 as a month: The limit returns the most recent sales, whatever their dates. Set start_date and end_date for a fixed window.
  • Taking a median over null prices: Count only rows with a sold_price, and keep the number of rows you dropped.
  • Using different filters on the 2 calls: Send the same dates and property_type to both, or the count stops describing the list.
  • Mixing property types: Without property_type, the list returns every type, from single_family to mobile. A median across them blends condos with single-family sales.
  • Asking for older sales: A date more than 365 days back returns a 400 that names the earliest date served. To compare with last year, store each month as you pull it.

Your account terms set where these outputs can be shown, so check them before any figure reaches your own customers.

General information. Not investment advice. Data as of 2026-11-07.

Related reading: How do I get foreclosure data for an app?, Is there a real estate listings API?.

Frequently asked questions

Is home sales data public record?
It depends on the state. Our comps endpoint docs warn that in a non-disclosure state, sale prices are not publicly recorded, and they name Missouri as one. In our sold-data rows, a withheld price comes back as a null sold_price.
Where can I find home sales data by ZIP code?
Call sold/count-by-zip for the number of sales in a ZIP code, at 1 credit. Then call sold/by-zip for one row per sale, at 1 credit per 20 properties returned. In one Denver ZIP code on 2026-11-07, the count call returned 173 single-family sales for 6 months.
How many homes sell in the US each year?
The National Association of Realtors put September 2026 existing-home sales at a seasonally adjusted annual rate of 3.99 million. That was down 2.2% from a year earlier. The median existing-home price was $417,500, up 1.3%.
Why do some sales have no price?
Our docs give private sales, off-market transfers and some agent sales as cases where the price is withheld. Those rows still appear, with sold_price set to null. In our Denver pull, all 40 rows carried a price.
How far back does the sold data go?
The 2 calls cover the trailing 365 days. A start_date or end_date older than that returns a 400 that names the earliest date we can serve, at 0 credits.

Try the call

Run sold/count-by-zip on one ZIP code first, then pull the list knowing what it costs. Parameters, fields and errors for both calls are in our API docs.

Read the API docs

Sources
Written by: The team at homesage.ai

We are a team of dedicated individuals with extensive experience in Real Estate, Home Improvement, and Artificial intelligence.  

Our mission is to help realtors, lenders, contractors and other professionals harness the power of AI to increase Business Volume.

Website Audit Icon Improve Performance

Increase Business Volume
with the power of AI

🔍

DealFinder Extension

Analyze any property listing instantly from your browser

Add to Chrome
📱

DealFinder Mobile App

AI-Powered Investment Analysis in Your Pocket

Scan QR to download DealFinder on the App Store App Store
Scan QR to download DealFinder on Google Play Google Play