Overview

Connect Claude, ChatGPT, and other AI assistants to the Model Match API via the Model Context Protocol.

The Model Match MCP server lets AI assistants query your mortgage, real estate, and originator data directly. Every API endpoint is automatically exposed as a tool — so Claude or any MCP-compatible client can search originators, pull market data, analyze loans, and more without writing code.

MCP (Model Context Protocol) is an open standard that lets AI applications connect to external data sources and tools. Learn more about MCP.

Connect your client

Pick your application to get setup instructions:

Most clients connect to the standard server URL:

https://mcp.modelmatch.com/mcp

ChatGPT uses a dedicated endpoint — see the ChatGPT guide.

How It Works

The MCP server dynamically generates tools from the Model Match OpenAPI specification. Every API endpoint becomes a callable tool with typed parameters — no manual configuration needed. When the API adds new endpoints, they're automatically available as tools.

Authentication uses OAuth — your AI client authenticates with your Model Match account, so tools respect your existing permissions and workspace access.

Available Tools

Tools are generated from every API endpoint. The tool name matches the endpoint's operationId from the OpenAPI spec (camelCase). Parameters are flattened into a single input schema — path parameters, query parameters, and request body fields all appear at the top level.

The server exposes a tool for every non-admin API endpoint. The complete, current list is always available via your MCP client's tool listing, or by inspecting the OpenAPI spec.

Example Tools

ToolDescription
listLoansSearch and filter loans with pagination
getLoanGet full loan details by ID
listOriginatorsSearch loan originators by name, NMLS, location
getOriginatorGet originator profile with production metrics
listAgentsSearch real estate agents
listCompaniesSearch mortgage companies
listPropertiesSearch properties with owner and valuation data
instantSearchCross-resource search across all entity types

All list tools support the same filtering, pagination, and sorting options as the REST API.

Tool names and parameters are derived from the OpenAPI spec and may change as the API evolves. Your MCP client will always show the current list of available tools.

Example Conversations

Once connected, you can ask your AI assistant questions like:

  • "Find the top 25 loan originators in Texas by volume"
  • "Show me Wells Fargo's branch production in California"
  • "What's the average loan size for purchase loans in Miami this year?"
  • "Look up originator NMLS #12345 and show their company history"
  • "Compare market activity in Austin vs Denver"

The assistant will call the appropriate Model Match tools, pass your filters, and present the results — all within the conversation.

Authentication Details

The MCP server uses OAuth 2.1 for authentication. When your AI client first connects:

  1. The client discovers the OAuth configuration via the /.well-known/oauth-protected-resource metadata endpoint (or /.well-known/oauth-protected-resource/mcp for clients that require the resource-suffixed path)
  2. You're redirected to Model Match to sign in and authorize access
  3. The client receives a token scoped to your account and workspace
  4. All subsequent tool calls use this token — your existing permissions apply

OAuth Scopes

The server requests the following scopes:

ScopePurpose
openidOIDC identity
profileBasic profile information
emailAccount email
offline_accessRefresh tokens for long-lived sessions

The server is its own OAuth 2.1 authorization server: it mints JWT access tokens and validates them via JWKS against its own keys — no token-introspection round-trip.

Rate Limits & Permissions

MCP tool calls are subject to the same rate limits and workspace permissions as direct API calls. If a tool returns a 403, your account doesn't have access to that resource.

On this page