MCP Server Wrapper for LLM Integration

Overview

This Node.js application serves as an MCP (Model Context Protocol) Server wrapper that integrates with Perplexity or other Large Language Models (LLMs) to provide inference capabilities for portfolio rebalancing. It runs inside a Docker container, ensuring easy deployment and scalability. This wrapper does not execute trades directly but instead supports generating portfolio rebalancing signals using state-of-the-art LLM-powered inference.

Features

• Runs as an MCP Server in a Docker container
• Connects to LLM inference APIs like Perplexity or any other supported LLM
• Provides portfolio rebalancing recommendations based on LLM-generated analytics
• No trade execution; the system focuses on signal generation and analytics
• Modular and extensible Node.js codebase for easy customization

Prerequisites

• Docker installed on your host machine
• Node.js (for local development and build)
• Perplexity desktop app or Claude
• Configuration of API keys or tokens for the robinhood api

Getting Started

Build and Run with Docker

1. Clone My repository:

2. Run the container:

   docker compose up

Note: Replace the environment variable values in the docker-compose.yml

References

• MCP Protocol and Server concepts
• Integration with Perplexity and other LLM APIs
• Portfolio rebalancing strategies with AI inference
• Perplexity Local MCP integration

Architecture

• Node.js MCP Server: Implements MCP Protocol for communication.
• LLM Integration Module: Handles calls to external LLMs such as Perplexity for inference.
• Docker Container: Provides isolated, reproducible run environment.

Development

Running Locally

1. Install dependencies:

   npm install

2. Start the application:

   node index.js

Testing

Run tests with:

npx @modelcontextprotocol/inspector node dist/index.js

Connecting a Locally Running MCP to Perplexity

Configure Perplexity to Connect to MCP

1. Go to below path Settings -> Connectors -> + Add Connector
2. Provide a Name "Robinhood" so your LLM targets the mcp
3. Under advanced tab

   Copy paste the perplexity.config.json content in Advanced tab in connectors

   Update your Perplexity environment variables to point to your MCP instance

4. Save and exit. You should be able to see your MCP showing runnng with number of tools.

Start Using and happy trading

Once connected, you can use Perplexity to send commands and queries to your local MCP seamlessly. and you should be able to see the mcp under sources

Note: Replace the environment variable values in perplexity.config.json