BUYER'S GUIDE
Agent-Ready CRE Data: APIs and MCP Servers for AI Agents (2026)
Updated October 2026 · 15 min read
TL;DR
For parcel and ownership data an assistant can call today, Regrid is the most agent-ready source: a public, documented MCP server on top of a self-serve API. ATTOM is the easiest institutional property dataset to try, with a 30-day API key. Cotality, Placer.ai, Prophia and Yardi Matrix offer official MCP access to customers, and Yardi Matrix and CRED iQ publish prices for it. RentCast, Parcl Labs and Realie publish self-serve prices with free tiers and MCP servers, though their data leans residential. CoStar and LoopNet offer no public API, and several vendors restrict sending data to third-party AI tools, so read the terms first.
AI assistants are now good enough to do real CRE research work: pull the parcels in a submarket, check who owns them, compare rents, read a lease portfolio, flag loans coming due. What stops them is access. An agent can only use data it can reach through a documented interface, under a licence that allows it. This guide covers the CRE data vendors that offer that access today, how you get it, what it costs where prices are published, and what the terms still forbid.
What MCP Is, in Plain Terms
The Model Context Protocol (MCP) is an open standard that gives AI apps one common way to connect to outside data and tools. Anthropic introduced it in November 2024, and in December 2025 donated it to the Agentic AI Foundation, a fund hosted by the Linux Foundation. A vendor that runs an MCP server publishes a set of tools, such as "look up a parcel by address" or "get rent comps", that any MCP-capable client can call: Claude, ChatGPT in developer mode, Cursor, VS Code or your own agent.
Why it matters to CRE teams:
- No export step. An analyst can ask a question in the assistant they already use and get an answer drawn from licensed data, instead of downloading a CSV and pasting it in.
- The vendor controls access. You sign in with your own account or key, so the agent sees only what your licence covers, and usage is metered like any other API call.
- One connection, many clients. A vendor builds one MCP server instead of a plugin for every AI app, which is why so many CRE data vendors announced one in 2026.
MCP is plumbing, not a licence. Access rights, prices and limits still come from your contract with the data vendor.
Four Kinds of Access
- MCP server: the most direct route for an AI assistant. Some are public with a setup guide; others are announced but provisioned only through sales.
- REST or GraphQL API: what a custom agent or developer calls. Most MCP servers sit on top of one.
- Data feed or bulk delivery: files or shares delivered to Snowflake, Databricks, S3 or SFTP. Best when an agent works over your own warehouse.
- Export only: CSV or Excel downloads from the app. An agent can read what you export, but nothing is live and terms may still restrict it.
Things to Know Before the List
- "Has an MCP server" can mean three things: a public server you can connect today with documentation (Regrid, RentCast, Parcl Labs), a server for existing customers with a setup guide (Prophia, Placer.ai, Cotality), or an announcement with access through sales (ATTOM, Trepp, CRED iQ Enterprise). The table below says which.
- Every agent call is a billed call. Regrid and RentCast bill MCP requests like API requests, and Placer.ai analyses consume credits. Agents loop, so set limits while you test.
- Read the AI clauses in the terms. Several vendors restrict sending their data to third-party AI tools or using it to train models. See "What agents still cannot do" below.
- Commercial depth is still mostly behind sales. Self-serve, published-price APIs lean residential (rents, home values, parcels). Institutional CRE data such as loans, comps and market research is available to agents, but mostly for existing customers.
Quick Comparison
| Tool | Data | Access for agents | How to get it | Published price |
|---|---|---|---|---|
| Regrid | Parcels, owners, zoning, buildings | MCP server, REST API, tiles, bulk files | Self-serve token; trial covers 7 counties | Overage $0.10 to $0.15 per record |
| ATTOM | Property, deed, mortgage, AVM | REST API, bulk, Snowflake, Databricks; MCP via sales | 30-day self-serve API trial | Not published |
| Cotality | Property, market, climate risk | MCP server, APIs, Snowflake, Databricks | Licensed customers | Not published |
| CRED iQ | CMBS, CLO, agency loans, public records | REST API; MCP and bulk feeds on Enterprise | Professional plan for API | $625 to $3,850/month; Enterprise custom |
| Placer.ai | Foot traffic, trade areas, visitors | MCP server, REST API, data feeds | API add-on; MCP per enterprise account | Not published |
| Yardi Matrix | Market rents, supply, ownership, loans | Claude connector; API and feeds via sales | Included on every Matrix plan | $199 to $1,999 per user per asset class per month |
| Prophia | Your own lease data | MCP server | Prophia customers | Not published |
| RentCast | Rents, comps, property records | MCP server, REST API | Self-serve key | Free 50 calls; $74 to $449/month |
| Cherre | Your unified portfolio data | GraphQL API, warehouse delivery | Sales | Not published |
| LightBox | Parcels, property, environmental | REST APIs, bulk, Snowflake | Free developer account | Not published |
| Parcl Labs | Housing market metrics | MCP server, REST API, CSV | Free account; MCP on Pro | Free 100 credits; Pro $50 per 4 weeks |
| Realie | Parcel and property records | MCP server, REST API, bulk | Self-serve | Free 25 tokens; $50 to $350/month |
| Plotzy | Parcels with zoning, owners | REST API | Scale plan via support | Scale plan custom |
| Trepp | CMBS and CRE loan data | MCP servers, data feed | Sales | Not published |
| Green Street | Research, valuations, REIT data | MCP server, API, Snowflake | Paid add-on for subscribers | Not published |
| MSCI Real Capital Analytics | Transactions and indexes | API, Snowflake, MCP connector | MSCI clients | Not published |
| RevolutionRE | Your multifamily PMS data | MCP server, data feeds | Included with subscriptions | BI Tool from $1.75 per unit per month |
| PropertyRadar | Owners, mortgages, distress | REST API, webhooks; MCP announced | Any paid plan | Plans $119 to $599/month |
| BatchData | Property, owner and contact data | REST API, bulk; MCP custom | Account with sandbox | Property data from $1,000/month |
| HelloData | Multifamily rents and concessions | REST API | Key on request | $0.10 per call (Portfolio plan) |
| Mashvisor | Short and long-term rental data | REST API | Self-serve | $129 to $599/month |
| Zoneomics | Zoning districts and rules | REST API | Self-serve | $134 to $549/month |
| SafeGraph | Points of interest, footprints | GraphQL API, bulk | Sales | Not published |
The 13 Best Agent-Ready CRE Data Sources
1. Regrid
Best for: Parcel, ownership and zoning lookups that any agent can call today
Regrid sells standardized parcel data across the US, Canada and parts of Europe, with building footprints, zoning and ownership updates as add-ons. It is the most agent-ready source in this guide: its MCP server is public and documented, with setup steps for Claude Code and Claude Desktop, and without a token it will answer questions about Regrid's schema, coverage and endpoints. With a token it queries parcel records, ownership, address and coordinate searches and county metadata through the same REST API that developers use.
Self-serve API plans include 2,000 records and 200,000 tiles a month, with overage published at $0.10 per record (Standard), $0.15 (Premium) and $0.001 per tile; the trial token covers seven counties, and larger volumes need an enterprise licence. Every parcel an agent returns counts against your quota, so Regrid suggests limiting results while you build. Its terms also require you to turn off model training and data sharing in your AI tool before using the data.
What's good
- Public MCP server with setup docs
- Same data by API, tiles and bulk files
- Published overage rates
- Works without a token for schema questions
What's not
- Base plan prices not on public pages
- Agents can burn quota fast
- Must disable AI training and data sharing
2. ATTOM
Best for: Property, deed, mortgage and valuation data for custom agents and data teams
ATTOM covers more than 160 million US properties (vendor claim) with characteristics, ownership, deeds and mortgages, foreclosures, permits, AVMs and risk data. Its REST API is documented publicly and anyone can get a 30-day evaluation key, which makes it the easiest institutional-grade property dataset to try with a custom agent. Bulk files and Snowflake and Databricks delivery suit teams that run agents over their own warehouse.
ATTOM has announced an official MCP server, but access is provisioned through the ATTOM Intelligence Platform via sales, with no public endpoint. API terms limit responses to 100 records, bar caching content for more than 24 hours and forbid using the data to build a database, so check the licence before letting an agent store results.
What's good
- Public API docs and a 30-day trial key
- Broad property, deed and mortgage coverage
- Snowflake and Databricks delivery
- Official MCP server announced
What's not
- No published API pricing
- MCP only through sales
- Strict caching and database-building limits
3. Cotality
Best for: Teams that already license Cotality (formerly CoreLogic) property data
Cotality runs an MCP server that gives AI agents access to its property data, market analytics and climate risk tools, with developer documentation and setup steps for VS Code, Claude Code, Cursor and Windsurf. The server only exposes the tools your token's scopes allow, so what an agent can do follows your licence. Data is also available through APIs and the Snowflake and Databricks marketplaces.
Access is for licensed customers, with credentials from your account team, and pricing is not published. If you already pay for Cotality data, this is one of the cleanest ways to put it in front of an assistant.
What's good
- Documented MCP server
- Tools scoped to your licence
- Snowflake and Databricks delivery
- Property, market and climate risk data
What's not
- Licensed customers only
- No published pricing
- Most tools need Cotality's property ID first
Access: MCP server, APIs, Snowflake, Databricks. MCP docs
Pricing: Not published.
4. CRED iQ
Best for: CRE debt data for agents at a published price
CRED iQ covers CMBS, SASB, CRE CLO and agency loans joined to public records, sales, ownership and comps, with a maturity pipeline, distress signals and valuation models. It is the clearest example of institutional CRE data with published prices for agent access: documented API access starts on the Professional plan at $3,850 a month, and the Enterprise plan adds an MCP server for Claude, ChatGPT and Cursor plus bulk feeds to Snowflake, Databricks and SFTP.
The catch is that the API and MCP documentation are not public, so you see them after you buy, and the Enterprise price is custom. The lower Solo and Starter plans include CSV and Excel exports only.
What's good
- Published prices for API access
- Loan-level debt data agents rarely get
- MCP server and bulk feeds on Enterprise
- Maturity pipeline and distress signals
What's not
- API docs not public
- MCP only on the custom-priced Enterprise plan
- Securitized and agency loans only for loan performance
Access: REST API (Professional), MCP server and bulk feeds (Enterprise). Plans
Pricing: Solo $625/month, Starter $1,250/month, Professional $3,850/month with API (billed annually); Enterprise custom.
5. Placer.ai
Best for: Retail, office and site teams that want foot traffic answers inside their assistant
Placer.ai's MCP server lets Claude, ChatGPT (developer mode), Copilot Studio agents, VS Code and Cursor run Placer analyses on foot traffic, trade areas, visitor demographics, cross-shopping and spend, returning an answer with a link to the full analysis. Sign-in uses your Placer account, and the email on your AI account must match. Separately, the PAPI REST API and data feeds to Azure, S3, GCS or Snowflake serve data teams.
MCP access is an enterprise entitlement granted per account and needs Placer's Pi assistant; it allows three concurrent analyses of up to ten minutes each and consumes Pi credits. The API is a paid add-on with weekly quotas. Plan prices are not published.
What's good
- Documented MCP setup for many clients
- Answers link back to the full analysis
- API and data feeds for data teams
- Free edition to explore the data
What's not
- MCP is an enterprise add-on
- No published pricing
- Concurrency and credit limits
6. Yardi Matrix
Best for: Market rents, supply, ownership and loan data in Claude at a published price
Yardi Matrix covers multifamily, affordable, student housing, single-family rental, office, industrial, self storage and land across US markets. Its Matrix Connector for Claude, built on MCP, lets the assistant query Matrix data, and the buy-now page says the connector is included on every plan. APIs and data feeds are available through sales.
Self-serve prices are published per user and per asset class: Starter $199, Professional $699, Expert $999 and Nationwide $1,999 a month. Multi-sector teams pay for each asset class, the connector currently targets Claude only, and the portal terms limit use to internal purposes with no bulk downloads.
What's good
- Claude connector included on every plan
- Published self-serve pricing
- Eight asset classes including office and industrial
- Ownership and loan intelligence
What's not
- Priced per user and per asset class
- Connector is Claude-only for now
- No bulk downloads under portal terms
Access: Claude connector (MCP); API and feeds via sales. AI connectors
Pricing: Starter $199, Professional $699, Expert $999, Nationwide $1,999 per user per asset class per month; Enterprise custom.
7. Prophia
Best for: Owners who want an assistant to answer questions across their own lease portfolio
Prophia abstracts commercial leases with AI and expert review, and its MCP connector lets ChatGPT, Claude or Gemini query that lease data: finding every lease tied to a tenant, checking insurance and compliance terms, reconciling CAM or modeling scenarios. A public setup guide covers Claude Desktop, Claude Code, Cursor and other clients, with sign-in through your organization's SSO or API keys.
This is your own data, not a market dataset, which avoids most licensing questions. It is only for Prophia customers, MCP must be enabled for the organization, and pricing is quote-based per tenant.
What's good
- Your own verified lease data in the assistant
- Public setup guide
- Works with ChatGPT, Claude and Gemini
- No third-party data licence to worry about
What's not
- Prophia customers only
- No published pricing
- Lease data only
Access: MCP server. Setup guide · Overview
Pricing: Quote-based, priced per tenant on an annual subscription.
8. RentCast
Best for: Rent estimates and rental comps by self-serve API and MCP
RentCast's API returns rent and value estimates, rental comps, property records and market statistics, and its official MCP server works with Cursor, VS Code, Claude and JetBrains tools. It is one of the few with fully published API prices and a free plan, and its API terms allow storing API data in your own systems and, within the permitted uses, reselling it, which is unusual.
API plans run from a free Developer plan (50 calls a month) to Foundation at $74, Growth at $199 and Scale at $449 a month, with per-call overage. MCP needs an active paid subscription, calls are billed like API calls, and the rate limit is 20 requests per second. Coverage is mainly residential and small multifamily.
What's good
- Published prices and a free plan
- Official MCP server
- Permissive API data terms
- Clear rate limits
What's not
- Residential and rental focus
- MCP needs a paid plan
- Thin on commercial assets
Access: MCP server, REST API. AI tools docs · API plans
Pricing: API Developer free (50 calls), Foundation $74, Growth $199, Scale $449 a month.
9. Cherre
Best for: Institutions that want agents working over their own unified portfolio data
Cherre connects, cleans and unifies data from internal systems and outside vendors into one model, and exposes it through a GraphQL API and managed delivery to Snowflake, AWS, Azure, BigQuery and SFTP. For a firm whose agent needs to reason across property management, accounting and third-party data at once, that unified layer is the hard part. RealPage completed its acquisition of Cherre in July 2026.
There is no MCP server and no public developer documentation, and access is through sales. Cherre's terms restrict using the services with third-party AI tools that move licensed data outside your licensed environment, and cap foundation data API calls at 100,000 a day.
What's good
- Unifies internal and third-party data
- GraphQL API and warehouse delivery
- Built for institutional data teams
- Agent tooling of its own
What's not
- No MCP server or public docs
- Enterprise pricing only
- Terms restrict third-party AI use
Access: GraphQL API, warehouse delivery. Platform
Pricing: Not published.
10. LightBox
Best for: Developers who want a broad catalogue of parcel, property and environmental APIs
LightBox publishes a developer portal with a catalogue of REST APIs covering parcels, property characteristics, building footprints and environmental data, plus bulk delivery and Snowflake. Anyone can open a free developer account and get an app key to start, which makes it easier to evaluate than most commercial data vendors.
There is no MCP server, trial quotas are not documented, an evaluation agreement applies, and pricing is not published.
What's good
- Public developer portal and API catalogue
- Free developer account
- Parcel, property and environmental data
- Snowflake delivery
What's not
- No MCP server
- No published pricing
- Evaluation terms apply
Access: REST APIs, bulk, Snowflake. Developer portal
Pricing: Not published.
11. Parcl Labs
Best for: Housing market metrics for agents on a small budget
Parcl Labs provides daily housing market data across US markets: prices, supply, sales, rentals and investor activity. Its AI Connect MCP server works with Claude, ChatGPT and Cursor, and the REST API and CSV downloader serve data teams.
A free account gets 100 credits a month; the Pro plan at $50 every four weeks includes 15,000 credits and AI Connect. Data is residential, the API licence limits use to internal purposes and requires attribution, and the vendor's own pages disagree about which paid plan includes MCP, so check before you buy.
What's good
- Free account and low-cost Pro plan
- Official MCP server
- Daily market metrics
- API and CSV access
What's not
- Residential only
- Attribution and internal-use licence
- Plan details differ across vendor pages
Access: MCP server, REST API, CSV. AI Connect · API docs
Pricing: Free (100 credits a month); Pro $50 every 4 weeks (15,000 credits).
12. Realie
Best for: Cheap, self-serve parcel records with an MCP server
Realie offers county-sourced records on more than 180 million US parcels (vendor claim) through a REST API, a CLI, bulk delivery to S3, GCS or Azure, and an official MCP server for Claude, Cursor and VS Code. Billing counts only successful calls, and the rate limit is 1,200 requests a minute.
Prices are published: a free tier with 25 tokens a month, then $50, $150 and $350 a month for 1,250, 6,000 and 30,000 tokens. A payment method is needed even for free API access. It is a newer vendor focused on property records rather than commercial leases, comps or debt.
What's good
- Published, low prices
- Official MCP server
- Bulk delivery to your cloud
- Only successful calls are billed
What's not
- Newer vendor
- Payment method needed for the free tier
- Property records, not CRE market data
13. Plotzy
Best for: Teams that want zoning-tagged parcel search in their own systems
Plotzy is a parcel and zoning research platform that tags every parcel with a two-level zoning designation and adds owner contacts, folders and notes. API access is part of its Scale plan, and the site publishes machine-readable files for AI agents (an llms.txt, an API catalogue and an agent skill file) that describe the product and how to authenticate.
There is no MCP server and no public API reference, API credentials are set up through Plotzy support rather than self-registration, and the Scale plan is custom-priced. For an agent that needs zoning-aware parcel search today, Regrid's zoning add-on or Zoneomics' self-serve zoning API are quicker to start with; Plotzy fits teams already using its app who want the same searches in their own tools.
What's good
- Zoning designation on every parcel
- Machine-readable files describe the product for agents
- Owner contacts alongside parcels
- 7-day free trial of the app
What's not
- API only on the custom Scale plan
- No MCP server or public API reference
- Still labeled beta
Also Documented for Agents
- Trepp offers TreppMCP servers for Claude and ChatGPT and a TreppData Feed, through sales (TreppMCP).
- Green Street sells a GreenStreetAI MCP server as a paid add-on, alongside its API, Excel add-in and Snowflake (GreenStreetAI).
- MSCI Real Capital Analytics delivers data by API and Snowflake, and its MSCI Connector for AI assistants is for MSCI clients; it does not say RCA transaction data is included (Data delivery).
- RevolutionRE includes a RevRE MCP server with its multifamily BI and ETL subscriptions; its public API documentation is listed as coming soon (RevRE MCP).
- PropertyRadar includes API access on every paid plan and says it offers an MCP server for Claude and ChatGPT (API).
- BatchData publishes data plans from $1,000 a month and a custom-priced MCP server, but its terms bar sending data to third-party AI systems without written approval (MCP).
- HelloData, Mashvisor and Zoneomics publish REST APIs for multifamily rents, rental returns and zoning (HelloData docs · Mashvisor API · Zoneomics API).
- SafeGraph and StreetLight Data offer documented APIs for points of interest and traffic, through sales (SafeGraph docs · StreetLight docs).
- Residential and early-stage options with official MCP servers include Zillapi (Zillow-sourced data from an unaffiliated vendor), Homesage.ai and Buildability (keyless, launched in 2026).
No Agent Access Today
- CoStar has no public API, and its terms of use bar exposing its products to outside AI tools, including retrieval-augmented generation, and sharing logins with AI agents.
- LoopNet has no data API; its LoopLink product embeds listings on broker websites.
- PropStream offers exports only, and its terms bar use of the service or its data by AI technology, including large language models.
What Agents Still Cannot Do With This Data
An MCP server removes the technical barrier. It does not remove the legal and practical ones.
- Licences still apply. Most data licences are internal use only and bar redistribution or building a database. ATTOM bars caching API content for more than 24 hours, SafeGraph says you may not cache its data, and Parcl Labs requires attribution. An agent that pastes results into a client memo or a shared knowledge base may breach that.
- Some vendors restrict AI directly. Regrid requires model training and data sharing to be off in your AI tool. BatchData bars sending data to third-party AI systems without written approval and bars using MCP outputs to train, embed or vectorize. Cherre restricts third-party AI tools that move licensed data outside your environment. MSCI's terms bar unauthorized AI agents and MCP connectors, with AI use covered by a contract supplement. CoStar and PropStream bar AI use outright.
- Rate limits and credits bite. RentCast allows 20 requests a second, HelloData 500 a minute, Realie 1,200 a minute, Placer.ai three concurrent analyses, and Cherre 100,000 foundation API calls a day. An agent exploring a question can make dozens of calls, each billed.
- The most valuable CRE data is the least open. Lease comps, private loan terms and deal pipelines sit with CoStar, CompStak, Trepp and similar vendors, mostly behind sales-led contracts or not available to agents at all.
- Answers still need checking. An agent can pick the wrong parcel, mis-join an owner or misread a field. Check numbers against the source record before they reach an investment committee or a lender.
For the wider landscape, see the Data & Analytics buyer's guide and the AI & Automation buyer's guide. For AI tools that read deal documents rather than call data, see the AI underwriting and AI lease abstraction guides.
How to Choose
- To give an assistant parcel and ownership data today: Regrid, with Realie as a cheaper self-serve option.
- For a custom agent over property, deed and mortgage data: ATTOM, starting with its 30-day trial key.
- If you already license the data: turn on Cotality, Placer.ai, Yardi Matrix, Trepp or Green Street access through your account team.
- For CRE debt data: CRED iQ, with API access from its Professional plan and MCP on Enterprise.
- For your own leases and portfolio data: Prophia for leases, Cherre or RevolutionRE for unified operating data.
- For rents on a budget: RentCast, or Parcl Labs for market-level housing metrics.
How We Researched This Guide
We worked from each vendor's own developer documentation, MCP pages, pricing pages and terms of use, checked in October 2026, and only included tools listed in this directory. We list access only where the vendor documents it publicly and link the page we read. We did not create API keys or connect these servers ourselves. Prices are the vendor's published rates; where a vendor does not publish prices we say so. Some vendor pages blocked automated checks, so those vendors (CompStak among them) are left out rather than described from secondhand reports. Terms are summarized, not quoted; read the full terms before connecting an agent. Rankings reflect how directly an agent can use the data and how relevant it is to CRE work, and no vendor paid for placement.
Vendor documentation, pricing and terms pages are linked in each entry above.
Sources
- "Introducing the Model Context Protocol", Anthropic, November 25, 2024
- "Donating the Model Context Protocol and establishing the Agentic AI Foundation", Anthropic, December 9, 2025
- What is the Model Context Protocol (MCP)?, modelcontextprotocol.io
Frequently Asked Questions
What is an MCP server for real estate data?
An MCP server is a vendor-run connector, built on the open Model Context Protocol, that lets AI assistants such as Claude or ChatGPT call the vendor's data tools directly, for example to look up a parcel or pull rent comps. You sign in with your own account or key, and calls are metered like API calls.
Which CRE data providers have an MCP server?
As of October 2026, Regrid, RentCast, Parcl Labs and Realie publish MCP servers with setup documentation; Cotality, Placer.ai, Prophia and Yardi Matrix offer them to customers; and ATTOM, Trepp, Green Street, CRED iQ (Enterprise) and RevolutionRE provide them through sales or subscriptions.
Does CoStar have an API?
CoStar does not publish a public API, and its terms of use bar exposing its products to outside AI tools. LoopNet, also owned by CoStar Group, has no data API.
Can I use property data with ChatGPT or Claude legally?
Only within your licence. Many vendors allow internal use but bar redistribution or caching, and some, such as BatchData, Cherre and MSCI, restrict third-party AI use specifically. Regrid requires model training to be turned off in your AI tool. Read the terms before connecting.
What is the cheapest way to give an AI agent CRE data?
Self-serve APIs with free tiers: RentCast (50 free calls a month), Parcl Labs (100 free credits a month) and Realie (25 free tokens a month) publish prices, and Regrid offers a trial token. These lean residential; commercial data such as loans and comps costs more and mostly goes through sales.
Browse All Data & Analytics Software
See every CRE data and analytics tool in our directory, with features, pricing and comparisons.
View All Data & Analytics Tools →