Why this list is short
The Model Context Protocol is an open standard Anthropic introduced in November 2024 for connecting AI applications to external systems. It is not a product, and nobody owns it.
That matters because the barrier is not the protocol. It is having data worth connecting and the licensing to let an assistant use it. Most property-data companies have the second problem.
We went looking for real estate MCP servers published by the company that owns the data, with a page or documentation you can read. We found four. Community wrappers around other vendors exist on GitHub. Still, a wrapper someone built over a public API isn’t the vendor offering a server, and it carries none of the vendor’s terms.
The real estate MCP servers you can actually connect to
1. Homesage.ai
The Homesage.ai MCP server exposes 33 tools over the same data our API serves, including the full 292-field property report, property condition grading, renovation cost estimates, rental estimates, and the Seller Motivation Score. You connect it from our dashboard with a personal link.
The useful part is not the tool count. It is that an assistant can chain them: resolve an address, grade the property condition, price the work, and project the return, in one conversation, without you writing the sequence.
Best for: Investors and developers who want an assistant to underwrite a deal end to end.
2. ATTOM
ATTOM publishes an MCP server described as schema-aware tools mapped to its data, covering property details, automated valuations, comparables, and sales history. Access is sales-led rather than self-serve: the page offers a trial request and a documentation link rather than an endpoint.
ATTOM is a records business at national scale. If the question you want an assistant to answer is about ownership, transactions or assessment history, that is the right pipe. It will not grade a property from photos.
Best for: Teams that need public records and valuation history at national scale.
3. Cotality
Cotality, formerly CoreLogic, launched an MCP server in March 2026, positioned as a universal connector between AI models and its data foundation. It sits alongside Trestle, the platform MLS organizations use to license listing data to brokers and technology providers.
There is no documented self-serve signup. Trestle access is MLS-gated by design: organizations join at no cost and approve connections by use case, and technology providers pay a monthly connection fee that is not published.
Best for: Brokerages and platforms already licensing MLS data through Trestle.
4. BatchData
BatchData says its MCP server is publicly accessible and can be connected to any MCP-compatible assistant, authenticated with a BatchData API key. Of the four, it is the most openly documented route to simply trying one.
The documentation is a vendor blog post rather than formal reference material, so check the current terms before building on it.
Best for: Developers who want to connect something quickly and see how MCP behaves.
How the four compare
| Server | What it is built on | How you get access | Published price |
|---|---|---|---|
| Homesage.ai | Our own property data and models | Dashboard link, self-serve | Credit-based, published |
| ATTOM | Public records and valuations | Request a trial, sales-led | Basic Property from $500/month |
| Cotality | Trestle MLS distribution | MLS-gated, by approved connection | Not published |
| BatchData | Property and contact data | API key, publicly accessible | Not published |
ATTOM’s $500/month is the only figure any of the four publishes for its data product; the rest quote on request. Read on each vendor’s own pricing page, 23 October 2026.
What an assistant can actually answer, measured
A connected assistant is only as good as the field coverage underneath it. So we measured what our own data carries across seven cities this week, using free preview calls on active MLS listings.
In Toledo, OH, 92.3% of 665 active listings carry a property condition grade and 84.7% carry a Seller Motivation Score. A rent estimate exists on 35.0%.
Property condition and negotiating room answer almost everywhere. Rent does not, and no protocol fixes that.
Across all seven cities, property condition landed between 86.9% and 94.0%, and the Seller Motivation Score between 82.3% and 90.6%. The rent estimate ranged from 10.8% to 35.4%, a spread of more than three times.
That shape decides what you should ask an assistant to do. A question about property condition or negotiating room will almost always get an answer. A question that depends on rent will get one in Dayton and often not in Columbus.
How to choose one
Start with the field your question depends on. An MCP server cannot invent a field its underlying data does not carry, and the four above are built on quite different foundations.
Then check how access works. Ours and BatchData’s can be connected directly. ATTOM’s runs through a trial request. Cotality’s depends on an MLS approving the connection, which takes weeks rather than minutes.
Finally, read what the assistant is permitted to do with what it receives. Listing data carries display and storage rules that don’t disappear just because an AI is holding it. Our guide to MCP servers covers the setup in more detail.
Estimated values, after-repair values, renovation costs and returns are model-derived estimates, not investment advice. Property data as of 23 October 2026.
Key takeaways
- MCP is an open standard from Anthropic, announced in November 2024. It is not a product and nobody owns it.
- Four property-data companies publish their own real estate MCP server today: ATTOM, Cotality, BatchData and Homesage.ai.
- Community wrappers around other vendors are not vendor offerings and carry none of the vendor’s terms.
- ATTOM publishes $500/month for Basic Property. The other three quote on request.
- Coverage decides what an assistant can answer: property condition and Seller Motivation Score are dependable across cities, rent is not.
Conclusion
The protocol is settled and open. What separates these four is the data underneath and how long it takes to get permission to use it. Pick the field your question needs, then choose how quickly you can connect. You can start now and connect ours from the dashboard.
Frequently asked questions
What is a Model Context Protocol server in real estate?
A server that exposes a property-data provider’s capabilities as tools an AI assistant can call directly. MCP is an open standard Anthropic introduced in November 2024 to connect AI applications to external systems.
Which real estate companies publish an MCP server?
As of 23 October 2026, four publish one on their own pages: ATTOM, Cotality, BatchData and Homesage.ai. Community-built wrappers around other APIs exist, but they are not vendor offerings.
Do I still need an API key?
Yes, in most cases. The server is a connection method, not a replacement for authentication or for your license to use the data.
Is there an official Zillow MCP server?
We couldn’t find one on Zillow’s own pages. The servers in circulation for Zillow data are third-party wrappers.
