Request Access
Data SamplesAPI DocsMCPMethodologyAboutRequest Access
Model Context Protocol

MCP Server

Connect AI agents directly to the intelligence layer. Query loans, resolve entities, and traverse the capital graph from Claude, GPT, or any MCP-compatible client.

Overview

What Is MCP

The Model Context Protocol is an open standard that lets AI agents call structured tools and read typed resources from external systems. Instead of pasting data into prompts or building custom integrations, MCP gives your agent direct, authenticated access to the full Real Intelligence data layer.

Structured Tools
Search loans, resolve entities, traverse the capital graph, and pull decision-ready signals. The agent picks the right tool for the query.
Schema-Defined Inputs
Every tool declares a typed input schema the agent reads at connection time. Field names, types, and constraints are validated before the call — no hallucinated parameters.
Authenticated Access
API-key scoped to your integration mandate. The server enforces the same access controls as the REST API. No data leakage across tenants.
Setup

Installation

Add the Real Intelligence MCP server to your client configuration. The example below shows Claude Desktop, but any MCP-compatible client works the same way.

claude_desktop_config.json
{
  "mcpServers": {
    "real-intelligence": {
      "command": "npx",
      "args": ["-y", "..."],
      "env": {
        "REAL_INTELLIGENCE_API_KEY": "ri_demo_key"
      }
    }
  }
}

// Package details provided with MCP access.

Replace ri_demo_key with the key issued in your onboarding package. The server authenticates on every call and scopes responses to your integration mandate.

Tools

Available Tools

A representative subset of the tool surface covering loan search, entity resolution, capital graph traversal, portfolio analytics, and signal retrieval. The full tool catalogue is provided on MCP access. The agent selects the right tool based on the user's natural-language query.

Loan SearchTool
Search and filter loan records across the full dataset. Accepts geographic, lender, balance, and maturity parameters with paginated results.
Example usage
"Find loans in New York above $15M maturing in the next 18 months"
Loan DetailTool
Retrieve the complete record for a single loan, including collateral context, debt sizing, and lender metadata.
Example usage
"Pull the full record for this loan"
Portfolio SummaryTool
Aggregate portfolio statistics across filters — total balance, loan counts, rate distribution, and pressure breakdown by state, lender, or property type.
Example usage
"Summarize the Texas office book across all tracked lenders"
Maturity WallTool
Return loans grouped by estimated maturity year, with aggregate balance and rate pressure context across the window.
Example usage
"Show the maturity wall for Florida multifamily between 2026 and 2028"
Entity LookupTool
Look up resolved entities by name or minimum confidence threshold. Returns relationship metadata and ownership context.
Example usage
"Find all resolved sponsor entities matching this name"
Ownership ChainTool
Walk the full ownership chain for a loan — from borrower through operating company to ultimate beneficial owner.
Example usage
"Who is the ultimate beneficial owner behind this borrower?"
Graph QueryTool
Traverse the capital graph from any node. Returns connected entities, relationship paths, and concentration metrics at configurable depth.
Example usage
"Map the capital chain from this lender's office portfolio to upstream allocators"
Public Debt SearchTool
Query the public debt layer across securitized and agency loan programs. Filter by geography, property type, balance, and maturity.
Example usage
"Find public-market loans in Atlanta over $40M maturing before 2027"
Signal FeedTool
Retrieve decision-ready flags for rate pressure, concentration risk, maturity timing, and other mandate-level filters.
Example usage
"Show high-severity signals in this market"
In Practice

Example Conversations

The MCP server turns natural-language questions into structured queries against the intelligence layer. Here is what that looks like in practice.

User prompt
“Show me loans in this market that are approaching maturity under pressure.”
What happens
The MCP server queries the loan dataset with the relevant filters and returns matching records with debt sizing, lender, and collateral context — structured for immediate analysis.
User prompt
“Who is the ultimate beneficial owner behind this borrower?”
What happens
The server chains multiple tools together — resolving the borrower entity, then walking the ownership chain through intermediate entities to the UBO, returning the full path.
User prompt
“Map the capital relationships for this lender's portfolio in this market.”
What happens
The server identifies the relevant positions, then traverses the capital graph to return connected entities, relationship paths, and concentration metrics.
Access

MCP Integration

MCP access is scoped alongside core infrastructure deployments. Same data layer, same access controls, delivered through the protocol your agents already speak.

Integration Credentials

MCP credentials are issued as part of engagement deployment. Contact the partnerships team for integration scope and access terms.

Includes authenticated server, full tool surface, schema-defined inputs, and usage aligned to engagement rate limits. Connect as many agents as the engagement requires.

Connect your agents

Structured CRE intelligence, native to your AI stack.

Request MCP access to receive server credentials and onboarding details.

Request MCP Access