Polyneural AI MCP Client

An MCP (Model Context Protocol) client bridge that converts HTTP MCP server requests to stdin/stdout JSON-RPC protocol for integration with Warp terminal and other MCP-compatible AI assistants.

Overview

This client acts as a bridge between:

• AI clients (like Warp terminal) that expect stdin/stdout JSON-RPC protocol
• Polyneural AI HTTP server that uses JSON-RPC over HTTP POST

Prerequisites

• Node.js v18.0.0 or higher
• A Polyneural AI API key

Installation

# Install dependencies (none required - uses built-in Node.js modules)
cd mcp-client
npm install

Configuration

The client is configured via environment variables:

Required Environment Variables

• POLYNEURAL_API_KEY - Your Polyneural.ai API key (format: kg_xxxxxxxx)

Optional Environment Variables

• POLYNEURAL_API_URL - API base URL (default: https://polyneural.ai/mcp)
• DEBUG - Enable debug logging (1 to enable, 0 or unset to disable)
• REQUEST_TIMEOUT_MS - HTTP request timeout in milliseconds (default: 30000)
• SHORT_TERM - Enable embedded short-term memory in tool descriptions (true/false, default: false)

Usage

Basic Usage

export POLYNEURAL_API_KEY=kg_your_api_key_here
node index.js

With Debug Logging

export POLYNEURAL_API_KEY=kg_your_api_key_here
export DEBUG=1
node index.js

Using npm scripts

# Basic usage
export POLYNEURAL_API_KEY=kg_your_api_key_here
npm start

# With debug logging
export POLYNEURAL_API_KEY=kg_your_api_key_here
npm run dev

Custom API URL

export POLYNEURAL_API_KEY=kg_your_api_key_here
export POLYNEURAL_API_URL=https://your-custom-api.example.com/mcp
node index.js

Integration with Warp Terminal

To use this client with Warp terminal, add it to your MCP configuration:

{
  "mcpServers": {
    "polyneural": {
      "command": "node",
      "args": ["/path/to/polyneural.ai/mcp-client/index.js"],
      "env": {
        "POLYNEURAL_API_KEY": "kg_your_api_key_here",
        "SHORT_TERM": "true"
      }
    }
  }
}

Integration with Claude Desktop

For Claude Desktop, add to your configuration file:

{
  "mcpServers": {
    "polyneural": {
      "command": "node",
      "args": ["/path/to/polyneural.ai/mcp-client/index.js"],
      "env": {
        "POLYNEURAL_API_KEY": "kg_your_api_key_here",
        "POLYNEURAL_API_URL": "https://polyneural.ai/mcp",
        "SHORT_TERM": "true"
      }
    }
  }
}

Supported MCP Methods

The client supports the following MCP protocol methods:

• initialize - Initialize the MCP connection
• tools/list - List available tools
• tools/call - Call a specific tool
• resources/list - List resources (returns empty list)
• resources/read - Read resources (returns error - not supported)
• prompts/list - List prompts (returns empty list)
• initialized - Initialization notification
• ping - Health check

Error Handling

The client includes comprehensive error handling:

• Connection errors - Returned as JSON-RPC error responses
• Timeout errors - Configurable timeout with graceful error responses
• Parse errors - Malformed JSON requests are handled gracefully
• Memory monitoring - Optional memory usage logging in debug mode

Logging

When DEBUG=1 is set, the client provides detailed logging including:

• Request/response sizes and previews
• Memory usage monitoring
• HTTP status codes and error details
• Message processing statistics

All debug logs are sent to stderr to avoid interfering with the JSON-RPC communication on stdout.

Protocol Details

Input (stdin)

• Expects newline-delimited JSON-RPC 2.0 requests
• Each line should be a complete JSON object
• Supports large message buffering

Output (stdout)

• Returns newline-delimited JSON-RPC 2.0 responses
• Each response corresponds to a request (except notifications)
• Includes proper error handling and ID matching

HTTP Communication

• Forwards requests to the configured API URL via HTTP POST
• Includes authentication via Bearer token
• Handles response mapping and error translation

Troubleshooting

Common Issues

1. "POLYNEURAL_API_KEY environment variable is required"

   • Make sure to set your API key: export POLYNEURAL_API_KEY=kg_your_key_here

2. Connection timeouts

   • Check your network connection
   • Verify the API URL is correct
   • Try increasing REQUEST_TIMEOUT_MS

3. Authentication errors

   • Verify your API key is correct
   • Check that your key has proper permissions

4. Large response handling

   • The client automatically handles large responses
   • Monitor memory usage with DEBUG=1

Debug Mode

Enable debug mode for detailed troubleshooting:

export DEBUG=1
export POLYNEURAL_API_KEY=kg_your_key_here
node index.js

This will show:

• All request/response details
• Memory usage statistics
• Error details and stack traces
• Message processing information

Development

Project Structure

mcp-client/
├── index.js          # Main MCP client bridge
├── package.json      # Project configuration
└── README.md         # This file

Making Changes

1. Edit index.js for client logic changes
2. Update package.json for dependency or metadata changes
3. Test with your MCP client (Warp, Claude Desktop, etc.)

Testing

Test the client manually:

export POLYNEURAL_API_KEY=kg_your_key_here
export DEBUG=1
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' | node index.js

License

MIT License - see the main project repository for details.