Your AI Platform Already Has the Wheel. Green Street Is the Engine.

The firms already using GreenStreetAI™ are completing in 15 minutes what used to take weeks. That is not a projection — it is what beta participants reported when they connected their AI platforms to Green Street’s MCP server for commercial real estate and started asking real questions against real data.

If your team is already using Claude, ChatGPT, Gemini, or other LLM tools for CRE work, the question is no longer whether AI belongs in your workflow. The question is what your AI is reasoning over — and whether that data is good enough to act on.

Your AI Is Only as Good as the Data It’s Reasoning Over.

Most AI tools in use today are trained on broad internet data. For general research, drafting, and synthesis, that is fine. For institutional CRE decisions — underwriting a market, benchmarking a transaction, evaluating a REIT — general knowledge is precisely where things go wrong.

“If your AI is reasoning over unverified data, you’re going to find very competent-sounding but wrong answers,” said Travis Valentine, Chief Technology Officer at Green Street. That gap — between an answer that sounds right and an answer that is verified — is the fault line that separates useful AI from a liability in a CRE context.

Green Street has spent over 40 years building proprietary research, market data, and analytics that institutional investors trust. The GreenStreetAI MCP server makes that intelligence available inside the AI tools your team already uses — without replacing your workflow, just upgrading what powers it.

Green Street: Slide 5 from the webinar deck – “Overview of MCP Server”.

What Is an MCP Server – and Why Does It Matter for CRE?

MCP stands for Model Context Protocol, an open standard originally developed by Anthropic. Think of it as a universal connector: instead of building a custom integration for every AI platform your firm might use, a single MCP server makes Green Street data available to any MCP-compatible client — Claude, ChatGPT, Gemini, and others.

The distinction that matters is between a traditional API and an MCP endpoint. An API is fixed — you ask for exactly one thing, you get exactly one thing back. MCP is different: it lets the AI model itself reason about which tools to call, how to call them, and in what order, to answer a multi-step question.

What happens between your query and the answer is more sophisticated than most users realize. When a question about our research reports, news articles, or learning center content hits the GreenStreetAI MCP server, the system doesn’t just search — it deconstructs. It parses the full conversation context, breaks the question into sub-questions, identifies entities — companies, markets, sectors — and uses those as filters before touching the research corpus. Then it pulls candidate chunks from across Green Street’s report library and runs a second AI evaluation pass, checking not just topical relevance but whether each chunk actually answers the specific question. A report can be about the right sector and still get cut because it misses a detail that changes the answer.

As Mike Wright, Senior Director of Artificial Intelligence at Green Street, put it: “We take that extra step to make sure we’re only taking into account the exact report chunks that actually truly answer your question.” This understanding of Green Street’s data — the sectors, the markets, the structure of the research — is built directly into the MCP server, so a firm that is new to the platform can still get to a precise answer without knowing how to ask a perfect query.

What Can You Actually Ask? Three Queries That Show the Difference.

The best way to understand what the GreenStreetAI MCP server makes possible is to see it applied to the kind of questions CRE professionals actually ask. The following examples are drawn from live demos — watch the webinar on-demand to see them in full.

Market underwriting – Orange County office

Ask a traditional API: “What is the cap rate for Orange County office?” It returns a number. Ask the same question through an MCP-connected model: “Underwriting an office in Orange County. What does Green Street have to color?”

The model identifies the relevant Green Street market overview tool, invokes it, and returns market data, market forecasts, recent sales comps, and a link to the market snapshot — in one response. No manual lookups and downloads across multiple tabs. No copy-paste from the Excel Add-In. No waiting for a data pull.

Cross-market data retrieval – Canadian retail

A.J. Apple, Senior Technical Product Manager for AI Initiatives at Green Street, ran a live query asking for Canadian retail market grades and cap rates. The model didn’t need to be told what sector ID Green Street uses for retail, or which markets fall under Canadian coverage. It used the available tools to look those up autonomously, then ran the data query with the correct parameters and returned a structured table of market grades and nominal cap rates across every Canadian retail market. The user typed a plain-language question. The model handled the data architecture. One of the advantages of using our MCP server over other data delivery methods is that our server and its tools have information to help guide the LLM in how to properly pull data.

Research synthesis — AI and data centers

Query: “Using Green Street research, how is AI affecting data centers?” Rather than returning a single report, the system broke the question down, identified relevant entities, pulled candidate chunks from across the research library, and ran an evaluation pass to confirm each chunk genuinely answered the question — not just that it was topically adjacent. The response came back with sourced findings and links to the underlying reports.

One query. Over twenty thousand reports and 150,000 news articles searched. A verified, attributed answer. That is not a search bar with better autocomplete. That is a reasoning layer sitting on top of over 40 years of proprietary CRE intelligence — available inside the AI tools your team is already using.

What Are Firms Already Doing With It? Here Are Some Results From the Beta.

Before the public launch, Green Street ran a structured beta program with client firms across different workflow types. The results were not theoretical — these were production workflows, run by analysts against live data.

Four themes emerged consistently:

• Weeks compressed to minutes. One multi-week, multi-market investment selection exercise was completed with a 15-minute prompt. Daily scheduled tasks for REIT and market analysis became automated routines.

• Green Street data inside existing tools. Beta participants pulled GS data directly into Excel to build sales comp tables. Others connected multiple MCP servers to power custom due-diligence workflows — without leaving the environments they already use.

• Conversational access to proprietary intelligence. Participants retrieved forecasts and property-level insights through plain-language queries, eliminating manual API calls and platform navigation.

• Trust in the underlying data. Beta participants consistently cited the quality and reliability of the research — calling it “phenomenal” — and specifically flagged it as a source they trusted for leadership-ready outputs.

That last point matters most. AI is only as reliable as the data it reasons over. The beta made clear that for institutional CRE professionals, data provenance is not a secondary consideration — it is the primary one. For more on why this distinction is shaping how firms evaluate AI tools, see Green Street’s piece on why trust is the new KPI in CRE AI.

What Happens to Your Data? A Straightforward Answer.

For any firm evaluating an AI data product, this is the question that matters before anything else: what happens to the queries, inputs, and outputs that flow through the system?

Green Street’s answer is unambiguous. Your data is never used to train, fine-tune, or improve any AI model. Authentication uses your existing platform credentials. Nothing new is stored. Nothing is shared.

For firms managing internal market intelligence, active deal pipelines, or client-sensitive analysis, this is not a checkbox — it is a fundamental requirement. When evaluating GreenStreetAI alongside other AI data tools, it is worth asking every vendor on your shortlist for the same commitment. The answer is not always the same.

What Is Available on the GreenStreetAI MCP Server Today?

As of August 2026, the following data sets are live and queryable through any MCP-compatible AI client:

GreenStreetAI CapabilityWhat It Gives You

Research Q&ALive · Beta

Ask questions across 20,000+ GS research reports — answers come back with source attribution and links to the underlying reports.

Market DataLive

Market grades, fundamentals, valuation, macro and demographics — queryable by market, sector, or region.

Sales CompsLive

Recent transaction comps with property-level detail.

ForecastsLive

Forward estimates across cap rates, CPPI, and occupancy — by sector and market.

AVM — Property ValuationLive

Automated valuation model for a single property, returned within the same conversation.

Company DataLive

Single-company metrics: rating, NAV premium, cap rate, target price.

AI Research SummariesLive

Condensed overviews of GS research reports across all regions — shareable at a glance.

Market OverviewsLive

Instant structured market brief: grades, fundamentals, recent comps for a given market and sector on demand.

GS News via MCPComing soon

Real-time CRE news integrated alongside structured data in a single conversation.

The MCP server is compatible with Claude, ChatGPT, Gemini, and other MCP-capable AI clients. Setup is handled by your organization’s admin — authentication uses existing Green Street credentials, and most connections take minutes rather than days. No new tooling. No engineering work on your side. Your AI environment stays the same; it simply gains access to better data — over forty years of independent CRE research, market analytics, company data, and transaction intelligence, available in every query.