MCP Server (Mortgage Comparison Platform)

MCP Server (Mortgage Comparison Platform)

By confersolutions GitHub

Canonical MCP server for parsing Loan Estimate (LE) and Closing Disclosure (CD) PDFs into MISMO-compliant JSON with LLM-enriched context. Built for AI-driven mortgage automation, compliance, and decisioning.

Overview

What is MCP Mortgage Server?

MCP Mortgage Server is a FastAPI-based server designed to parse Loan Estimate (LE) and Closing Disclosure (CD) PDFs into MISMO-compliant JSON format, enriched with LLM context for AI-driven mortgage automation and compliance.

How to use MCP Mortgage Server?

To use the MCP Mortgage Server, clone the repository, set up a virtual environment, install dependencies, configure environment variables, and run the server using Uvicorn. You can also deploy it using Docker.

Key features of MCP Mortgage Server?

  • Parses Loan Estimate (LE) and Closing Disclosure (CD) PDFs into MISMO-compliant JSON.
  • Built-in LLM context generation for compliance checks.
  • FastAPI-based REST API with OpenAPI documentation.
  • Docker support for easy deployment.
  • Includes testing capabilities with pytest.

Use cases of MCP Mortgage Server?

  1. Automating the parsing of mortgage documents for compliance.
  2. Integrating with AI tools for enhanced decision-making in mortgage processing.
  3. Streamlining the mortgage application process by converting PDFs to structured data.

FAQ from MCP Mortgage Server?

  • Can MCP Mortgage Server handle all types of mortgage documents?

Yes, it is specifically designed to parse Loan Estimates and Closing Disclosures.

  • Is there a demo available?

Yes, you can run the server locally and access the OpenAPI documentation for testing.

  • What technologies are used in MCP Mortgage Server?

The server is built using Python with FastAPI and supports Docker for deployment.

Content

MCP Server (Mortgage Comparison Platform)

A FastAPI-based server that provides mortgage document parsing and comparison tools through a standardized API. The server is designed to be easily integrated with various AI frameworks including CrewAI, AutoGen, and LangChain.

Currently implements a basic "hello" tool as a proof of concept, with mortgage document parsing tools coming soon.

Version License: MIT Website

Status

This is a beta release (v0.1.0) that provides:

  • Core server infrastructure with security features
  • Basic "hello" tool for testing framework integrations
  • Example integrations with CrewAI, AutoGen, and LangChain

Future versions will add mortgage document parsing and comparison tools.

Features

  • FastAPI server with production-ready features:
    • API key authentication
    • Rate limiting support
    • CORS middleware configuration
  • Framework integrations for AI agents:
    • CrewAI
    • AutoGen
    • LangChain
  • Extensible architecture for adding mortgage parsing tools
  • Open source for transparency and community contributions

Quick Start

  1. Clone the repository:
    git clone https://github.com/confersolutions/mcp-mortgage-server.git
    cd mcp-mortgage-server
    

Roadmap

  • ✅ Core server infrastructure with security and rate limiting
  • ✅ Framework integrations (CrewAI, AutoGen, LangChain)
  • ✅ Basic tool implementation ("hello" endpoint)
  • 🚧 Loan Estimate (LE) parsing to MISMO format
  • 🚧 Closing Disclosure (CD) parsing
  • 🚧 Mortgage comparison tools
  • 🚧 Additional mortgage document analysis features

Installation

  1. Clone the repository
  2. Create a virtual environment:
    python -m venv .venv
    source .venv/bin/activate  # On Windows: .venv\Scripts\activate
    
  3. Install dependencies:
    pip install fastapi uvicorn slowapi python-dotenv
    pip install crewai autogen langchain langchain-openai
    

Configuration

Create a .env file in the root directory with the following variables:

API_KEY=your_api_key_here
RATE_LIMIT_PER_MINUTE=120
ALLOWED_ORIGINS=http://localhost:3000
HOST=0.0.0.0
PORT=8001
WORKERS=1

Running the Server

python server.py

The server will start on http://localhost:8001 by default.

API Endpoints

Health Check

GET /health
Response: {"status": "healthy"}

List Available Tools

GET /tools
Headers: X-API-Key: your_api_key_here
Response: List of available tools and their configurations

Call Tool

POST /call
Headers: X-API-Key: your_api_key_here
Body: {
    "tool": "hello",
    "input": {
        "name": "World"  // Optional
    }
}
Response: {
    "output": "Hello, World!"
}

Framework Integration Examples

See examples/test_all_integrations.py for examples of how to use the server with:

  • CrewAI
  • AutoGen
  • LangChain

CrewAI Example

from crewai import Agent, Task, Crew
from mcp_toolkit import MCPToolkitCrewAI

toolkit = MCPToolkitCrewAI()
tools = await toolkit.get_tools()

agent = Agent(
    role="Greeter",
    goal="Say hello to the user",
    tools=tools
)

task = Task(
    description="Say hello to the user",
    agent=agent
)

crew = Crew(
    agents=[agent],
    tasks=[task]
)

result = await crew.kickoff()

AutoGen Example

from autogen import AssistantAgent, UserProxyAgent
from mcp_toolkit import MCPToolkitAutoGen

toolkit = MCPToolkitAutoGen()
tools = await toolkit.get_tools()

assistant = AssistantAgent(
    name="assistant",
    llm_config={"tools": tools}
)

user_proxy = UserProxyAgent(
    name="user_proxy",
    code_execution_config={"use_docker": False}
)

await user_proxy.initiate_chat(assistant, message="Please say hello to Alice")

LangChain Example

from langchain.agents import Tool, AgentExecutor, create_react_agent
from langchain_openai import ChatOpenAI
from mcp_toolkit import MCPToolkitLangChain

toolkit = MCPToolkitLangChain()
tools = [
    Tool(
        name="hello",
        func=lambda x: asyncio.get_event_loop().run_until_complete(toolkit.hello(name=x)),
        description="A tool that says hello to someone",
        return_direct=True
    )
]

llm = ChatOpenAI(temperature=0)
agent = create_react_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools)

result = await agent_executor.ainvoke({"input": "Please say hello to Bob"})

Rate Limiting

The server implements rate limiting using slowapi. By default, it's set to 120 requests per minute per IP address. This can be configured using the RATE_LIMIT_PER_MINUTE environment variable.

Security

  • API key authentication is required for all endpoints except /health
  • CORS is configured to allow specific origins (set via ALLOWED_ORIGINS environment variable)
  • All exceptions are caught and returned with appropriate error messages

Contributing

Feel free to open issues or submit pull requests for improvements.

About

This project is maintained by Confer Solutions. For questions or support, contact us at info@confersolutions.ai.

License

MIT License - see LICENSE file for details.

No tools information available.
No content found.