A Zillow MCP server is a Model Context Protocol server that gives an AI agent such as Claude, ChatGPT or Cursor tools for Zillow data: homes for sale or rent in an area, with price, beds, baths, living area, Zestimate, rent Zestimate and days on Zillow. For research, comps and anything you want in a spreadsheet, the best Zillow MCP setup is the hosted Scrape.ly MCP server: your agent runs ready-made Zillow scrapers in the cloud and gets every listing as a structured row it can calculate with, in any MCP client, with no Bridge API access to request.
This guide explains how that differs from Zillow's own new connector, shows the setup per client, the Actors and prices, the tool calls, and real runs: five Boise homes with Zestimates and rent estimates in 5.1 seconds for under a cent, then the detail fields the agent pulls next. For short-term rentals, see Airbnb MCP.
Two kinds of Zillow MCP server
Zillow now has its own. Since 19 October 2026 there's a Zillow connector in Claude, and Zillow writes that its ChatGPT app and the Claude connector "use the same Zillow Model Context Protocol (MCP) connection to search listings, explore home values and estimate what someone can afford." You turn it on under Customize → Connectors, no Zillow account needed.
It's built for shopping for a home. Zillow says it made the connector "to make its MLS-approved listing display and housing tools available in the chatbot", that "Zillow controls the data and tools available through the connector, while Claude generates the conversational response", and that tours, agents and financing continue on Zillow. Of the ChatGPT app, which runs on the same connection, Zillow adds that "AI models are not allowed to train on or have access to licensed data."
That's a good way to browse homes in a chat. It's a different job from research, where the agent needs the listing data itself: forty listings in a table, price per square foot, asking price against the Zestimate, a gross rent yield, a CSV for a model, the same search every Monday. The rest of page one offers that through someone's API: sap156/zillow-mcp-server, the top GitHub result, requires "A Zillow Bridge API key" that you request access for, and other servers resell Zillow-sourced data through their own credit-based APIs.
The Scrape.ly server reads public Zillow pages instead, and hands the agent rows.
| Scrape.ly MCP server | Zillow's connector | Bridge API servers | |
|---|---|---|---|
| Built for | Research, comps, exports, monitoring | Home shopping in Claude and ChatGPT | Apps with approved API access |
| What the agent gets | One row per listing: price, beds, baths, area, Zestimate, rent Zestimate, tax-assessed value, days on Zillow | Zillow's listing display, home values and maps in the chat | Whatever the API returns |
| Access | A Scrape.ly account, OAuth sign-in | Connect in Claude; ChatGPT app | A Bridge API key you request |
| Clients | Claude, ChatGPT, Cursor, Claude Code, VS Code, scripts | Claude (web, desktop, mobile) and ChatGPT | Depends on the server |
| Export and schedule | JSON, CSV or Excel datasets; schedules and automations | Next steps continue on Zillow | Depends on the server |
How to connect the Scrape.ly MCP server
The Scrape.ly MCP server is hosted at https://mcp.scrape.ly, speaks MCP's streamable HTTP transport and signs you in with OAuth or a Scrape.ly API token. Create a free account first; the free plan includes $5 of usage every month, no card needed.
Claude Code. One command, then /mcp inside Claude Code to sign in:
claude mcp add --transport http scrapely https://mcp.scrape.lyClaude Desktop and claude.ai. Add https://mcp.scrape.ly as a custom connector under Customize → Connectors; the Claude MCP guide has the steps. If Zillow's connector is on too, keep in mind that Zillow says naming "Zillow" in a prompt makes Claude more likely to use its connector, so say "use scrapely" when you want the data.
Cursor. Add this to ~/.cursor/mcp.json, or leave out headers to sign in with OAuth:
{
"mcpServers": {
"scrapely": {
"url": "https://mcp.scrape.ly",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}ChatGPT, VS Code, Windsurf and others use the same URL; formats are in the MCP docs.
A Zillow-only server. To give the agent just the two Zillow tools:
https://mcp.scrape.ly?actors=roamandroof/zillow-search-scraper,roamandroof/zillow-detail-scraperThe Zillow scrapers your agent can use
| Actor | Input | What comes back | Price per 1,000 |
|---|---|---|---|
| Zillow Search Scraper | City, ZIP or search URLs (filters from the address bar work) | Price, status, home type, beds, baths, living area, lot size, Zestimate, rent Zestimate, tax-assessed value, days on Zillow, photos, listing URL | $1.51 listings |
| Zillow Detail Scraper | Home URLs, street addresses or a location search | Everything above plus year built, price per square foot, HOA fee, tax rate, description, MLS name and ID, broker, structured property facts | $1.31 listings |
Both are pay per listing with platform usage included, as listed on their Store pages on 26 October 2026. A URL or address that can't be read saves no row and isn't charged. There's also a Zillow Agents Scraper ($1.61 per 1,000 agents), but it returns names, phone numbers and emails of people, so use it only with a lawful reason. The Zillow scrapers page compares all three with sample output.
What the agent does: the four tool calls
search-actorswith "zillow" finds the Actors above.fetch-actor-detailsreturns the input schema, README and price, so the agent knows a city page goes insearchUrlsand the cap inresultsLimit.call-actorstarts the run and waits up to 45 seconds; both runs here finished in about five.get-actor-outputreads the rows, only the fields the agent asks for.
The rows never land in the call-actor reply, so an 800-listing export doesn't flood the context window.
A real run: five Boise listings with Zestimates
The prompt: "Use scrapely to get five homes for sale in Boise, ID and compare each asking price with its Zestimate and its rent Zestimate."
The agent sends roamandroof/zillow-search-scraper the city page, with a cap:
{
"actor": "roamandroof/zillow-search-scraper",
"input": {
"searchUrls": [{ "url": "https://www.zillow.com/boise-id/" }],
"resultsLimit": 5,
"extractionMethod": "PAGINATION"
},
"waitSecs": 45,
"callOptions": { "maxTotalChargeUsd": 0.05 }
}The run on 26 October 2026 took 5.1 seconds. Its log:
2026-10-26T14:25:52.355Z [scrapely.actor] INFO Searching 1 URL(s) with PAGINATION, limit 5, proxy DATACENTER
2026-10-26T14:25:53.959Z [scrapely.actor] INFO Finished with 5 listings, 1 requests, 672899 bytes.The agent read the fields it needed:
{
"runId": "c957edf8113b49428",
"fields": "listingAddress.zipCode,listingPrice,bedrooms,bathrooms,livingArea,zestimate,rentZestimate,daysOnZillow"
}These are private homes, so we show the ZIP code only, never the street:
| ZIP | Asking price | Beds / baths | Sq ft | $ per sq ft | Zestimate | Asking vs Zestimate | Rent Zestimate | Days on Zillow |
|---|---|---|---|---|---|---|---|---|
| 83704 | $449,990 | 4 / 2 | 1,822 | $247 | $445,700 | +1.0% | $2,159 | 3 |
| 83703 | $950,000 | 4 / 3 | 2,848 | $334 | $953,400 | −0.4% | $4,275 | 5 |
| 83705 | $469,000 | 3 / 1 | 1,040 | $451 | $463,600 | +1.2% | $1,905 | 17 |
| 83703 | $998,000 | 5 / 3 | 3,589 | $278 | $977,200 | +2.1% | $4,441 | 44 |
| 83703 | $1,050,000 | 6 / 3 | 3,777 | $278 | $1,015,300 | +3.4% | $5,431 | 4 |
The last three columns of math are the agent's, computed from the rows. Four of five homes were listed slightly above their Zestimate, by up to 3.4%. As rentals, the gross yield (rent Zestimate × 12 ÷ price) ran from 4.9% for the small 83705 house to 6.2% for the six-bedroom in 83703. Each row also carried the lot size, tax-assessed value, photo count, the listing brokerage and flags such as for-sale-by-agent.
Cost: 5 × $0.00151 = $0.00755 at list price, platform usage included. Our account was billed $0.00604 after its plan's Store discount.
The next call: details for the homes that matter
Search rows are the overview. When the agent needs year built, price per square foot, HOA fees or the MLS name, it passes the listing URLs from the search to the Detail Scraper's startUrls:
{
"startUrls": [
{ "url": "https://www.zillow.com/homedetails/…_zpid/" },
{ "url": "https://www.zillow.com/homedetails/…_zpid/" }
],
"maxItems": 2
}We've shortened the URLs because they contain street addresses. The Detail Scraper also takes a city directly. Here's what a run of {"searchQueries": ["Austin, TX"], "maxItems": 5, "propertyStatus": "FOR_SALE"} returned on 25 October 2026, in 5.5 seconds:
| ZIP | Price | Beds / baths | Sq ft | $ per sq ft | Year built |
|---|---|---|---|---|---|
| 78747 | $449,000 | 4 / 3 | 2,428 | $185 | 2017 |
| 78739 | $1,375,000 | 5 / 4 | 4,395 | $313 | 2004 |
| 78729 | $660,000 | 5 / 4 | 3,154 | $209 | 1995 |
| 78702 | $995,000 | 3 / 3 | 1,530 | $650 | 1900 |
| 78750 | $1,275,000 | 4 / 5 | 4,073 | $313 | 1985 |
The 1900 house in 78702 cost more than three times as much per square foot as the 2017 build in 78747, the kind of spread an agent should flag before averaging. Five listings cost 5 × $0.00131 = $0.00655 at list price. Our Zillow data tutorial walks through every detail field and the Python client.
Prompts that work well
- "Use scrapely to get 40 homes for sale in ZIP 78704 and give me the median price per square foot by bedroom count."
- "Search Zillow rentals in Denver under $2,500 a month with the Zillow Search Scraper, 30 listings, and list the five with the most living area."
- "Every Monday, pull new listings for this search URL and tell me which ones are priced below their Zestimate."
- "Get details for these three listing URLs and compare HOA fees and year built."
- "Keep every run under $0.10."
The agent turns the last line into callOptions.maxTotalChargeUsd.
What your agent can and can't get from Zillow
- Public pages only. Listings behind a login and saved searches are out of scope.
- Search depth. Plain pagination stops at the 20 pages Zillow shows for one search, about 800 listings; the default zoom-in mode splits a dense map to go further. The Detail Scraper's own location search returns the first results page, about 40 homes.
- No price or tax history. The Detail Scraper doesn't fill price history, tax history, nearby schools or nearby homes, per its README. Current price, Zestimate, tax-assessed value and tax rate are there.
- Fields Zillow doesn't show stay empty. Agent contact details or days on Zillow are missing on some listings; the Actors return
nullinstead of guessing. - Addresses point to people's homes. Treat street addresses and agent names as personal data. We withhold both in this article, and you should keep them out of anything you publish.
Is it legal to use Zillow data with an AI agent?
The Actors read publicly visible listing pages and never sign in. Zillow's Terms of Use, updated October 28, 2025, ask users not to "conduct automated queries (including screen and database scraping, spiders, robots, crawlers...)", and allow copying "without the aid of any automated processes" for personal or Pro use. How far a site's terms bind logged-out access is a legal question courts have looked at; our guide to web scraping and the law covers the main cases. Collect only what you need, keep personal data out, and get legal advice before commercial use. This isn't legal advice.
Spending control for agents
- A dollar cap per run.
maxTotalChargeUsdcaps everything a run can cost. - Scoped API keys with daily charge limits. A scheduled agent gets its own key under Settings → API keys with only
actors:runandruns:read. - A prepaid balance. No invoice surprises.
For a REST endpoint instead of a chat, the Zillow API page has Python, JavaScript and cURL examples. For the wider picture of property data, see real estate data scraping; the same server covers other sites too, as in Google Maps MCP.
FAQ
Does Zillow have an MCP server?
Yes. Since 19 October 2026 Zillow offers a connector in Claude built on MCP, and the same connection runs its ChatGPT app. It shows Zillow's listing display, home values and affordability tools in the chat. For listing data your agent can analyze and export, in any MCP client, use the Scrape.ly MCP server.
What is the Zillow MCP server URL for Scrape.ly?
https://mcp.scrape.ly, for Claude, ChatGPT, Cursor and every other client. Add ?actors=roamandroof/zillow-search-scraper,roamandroof/zillow-detail-scraper to expose only the Zillow tools.
Does Zillow have a free API?
The top open-source Zillow MCP server runs on the Zillow Bridge API, which needs an API key you request access for. The Scrape.ly Actors need no Zillow API access: you pay $1.31 to $1.51 per 1,000 listings, and the free plan's $5 a month covers over 3,000 search listings.
Does Zillow allow data scraping?
Zillow's Terms of Use say users shouldn't run automated queries, including scraping, against its services. The Scrape.ly Actors only read publicly visible pages without signing in, but the terms still apply to your use. Read them, keep personal data out, and get legal advice for commercial projects.
Can I use Zillow MCP with Claude?
Yes, two ways. Zillow's own connector shows listings for home shopping. The Scrape.ly MCP server, added as a custom connector or with claude mcp add in Claude Code, gives Claude the listing data itself, so it can compare prices, Zestimates and rents across a whole search.
How much does a Zillow MCP setup cost?
Connecting the Scrape.ly server is free. You pay per listing: $1.51 per 1,000 from the Search Scraper and $1.31 per 1,000 from the Detail Scraper, platform usage included. Our five-listing Boise run cost $0.00755 at list price.
