What is the Rabbet MCP? A Guide to AI in Real Estate Development Finance
A guide for real estate developers and construction lenders on what the Rabbet MCP connector does, how it keeps data safe, and how teams are using it.
You may have heard a new term recently: MCP, short for Model Context Protocol. It sounds technical, but the idea is simple. It is a standard way for the AI tools you already use to reach the systems where your work lives, with your permission.
Rabbet has built one. You can ask your AI assistant questions about your projects, portfolio, draws, and budgets and get answers pulled straight from Rabbet Development and Rabbet Construction Lending in seconds.
Whether you are a real estate developer managing active projects or a construction lender reviewing hundreds of loans, this guide answers the most common questions about how it works and how it helps you.
The basics
What exactly is an MCP?
MCP stands for Model Context Protocol. Think of it as a bridge between your AI assistant and a software system like Rabbet. Before MCP, if you wanted AI help with your real estate development finance data, you had to copy information out of Rabbet and paste it into a chat. With an MCP connection, your AI assistant reads your Rabbet data directly and answers from it.
Think of MCP like a standard electrical plug. Any lamp with a standard plug works in any outlet in your house. MCP gives AI tools a standard, secure way to plug into software systems.
What is the Rabbet MCP connector?
It is a secure connection that lets your AI assistant read information from Rabbet Development or Rabbet Construction Lending. Once it is set up, you ask questions in plain language and get answers drawn from your structured financial data.
What can it read?
Budgets, draws and draw history, funding sources, vendors, checklist results, and document status. It reads the same structured data your team works with in Rabbet every day.
Which AI tools work with it?
Any AI tool that supports the Model Context Protocol standard with OAuth 2.1 authentication. This includes:
- Claude: available on desktop, web, and mobile.
- ChatGPT: requires enabling developer mode in settings during the one-time setup.
- Other MCP-compatible tools: any tool supporting MCP with OAuth 2.1 can be used, including Microsoft Copilot.
Do I need to be technical to use this?
No. Setup is a one-time process, and in most organizations it is handled for you. If you use Claude, ChatGPT, or Copilot on a Team, Business, or Enterprise account, your AI workspace administrator adds the Rabbet connection once and makes it available to your team. On an individual account, you add it yourself in your AI tool's settings.
Either way, once it is set up you interact in plain language, the same way you would message a colleague. You do not need to learn prompt engineering or write code.
How it works
How do I actually use it?
- Sign in to Rabbet. The first time you use the Rabbet connection, your AI assistant asks you to sign in with your own Rabbet login and approve access. You use your own credentials, which is what ties every answer to the permissions you already have.
- Ask a question. Open your AI assistant and ask about your budgets, draws, loans, projects, vendors, or documents.
- Get a grounded answer. The assistant reads what it needs from Rabbet and answers from your actual data.
What can it see in Rabbet?
Only what you already have access to, and only after your organization has turned it on for you. Access is controlled in three layers:
- Organization enables MCP. An administrator turns MCP on for your organization.
- Admin grants user access. An administrator grants access to specific users.
- Scoped questions. Every question is scoped to your user permissions. You can only ask about data you are already permitted to see in Rabbet.
Can it change things in Rabbet?
No. It is strictly read-only. It can look things up and summarize them, but it cannot create, edit, delete, approve, or send anything on your behalf.
What kinds of questions work best?
Questions grounded in the data Rabbet tracks.
For real estate developers (Rabbet Development):
- "What is my remaining contingency on the Westside project, and how much has been used by division?"
- "Which line items are billing ahead of overall project progress?"
- "What is drawn against each funding source, and is anything close to its commitment?"
- "Which vendors have we paid without a lien waiver on file?"
For construction lenders (Rabbet Construction Lending):
- "Which projects in my portfolio have a division more than 10 percent over budget?"
- "Compare the last two draws on this project and flag what changed."
- "What is blocking this draw from advancing?"
- "Summarize this draw and call out anything unusual."
Generating reports and summaries
You can also ask your AI assistant to put answers into a format you can share, such as a one-page project status summary or a table of active projects with their latest draw amounts. The assistant reads the data from Rabbet and formats it the way you ask.
Security and your data
How does the connection stay secure?
It uses the same authentication standards that protect Rabbet today. You sign in through Rabbet's normal login flow and grant permission explicitly. All traffic between your AI assistant and Rabbet is encrypted, and access is tied to your user permissions.
Can I turn it off?
Yes. You can manage and remove MCP connections in the settings of your AI tool at any time. A Rabbet administrator can also revoke your access in Rabbet.
Is my data being used to train AI models?
Rabbet does not use your data to train AI models. On your AI provider's side, training policies vary between providers such as Anthropic and OpenAI, and between their business plans. Most business and enterprise AI plans include commitments not to train on customer data. We recommend checking your provider's data policies to confirm.
What about AI hallucinations?
When an AI assistant answers from its own general knowledge, it can guess. When it answers through the Rabbet MCP connector, it reads actual structured data from your account, which substantially reduces the chance of a wrong answer. For any decision that affects compliance or funding, we still recommend verifying the final numbers in Rabbet directly.
What teams are doing with it today
- Draw and budget questions on demand. Contingency burn, budget variances, stalled draws, answered in seconds.
- Faster portfolio reviews. Ask for a portfolio summary to spot trends, flag outliers, and surface projects that need attention.
- One-off analysis without the export. Questions like average draw turnaround that would normally mean pulling data into a spreadsheet.
- Preparing for meetings. Before a call with a partner, borrower, or internal team, ask for a summary of the relevant project so you walk in with the facts.
How to get it
The Rabbet MCP connector is available now to customers on the Premium tier. Reach out to our support team or your customer success manager to turn it on. Not a customer? The reason this AI integration works is that the data underneath it is already structured. See how Rabbet works.
