PolyNeural.ai MCP Bundle (@polyneural/mcpb)

An MCP Bundle that provides AI agents with persistent memory through the PolyNeural.ai knowledge graph platform.

🚀 Quick Download

Download Latest Release →

Double-click the .mcpb file to install in Claude Desktop, then configure your PolyNeural.ai API key.

Architecture

This MCP Bundle implements a stdio MCP server that bridges to the PolyNeural.ai backend HTTP MCP endpoints. This approach provides:

• ✅ No code duplication - Uses existing PolyNeural.ai backend MCP endpoints
• ✅ Seamless authentication - Forwards API keys via HTTP headers
• ✅ Protocol translation - Bridges stdio MCP ↔ HTTP MCP endpoints
• ✅ Easy maintenance - Single source of truth for MCP tools and logic

Prerequisites

1. PolyNeural.ai Backend Running: The backend must be accessible (default: https://polyneural.ai)
2. Valid API Key: You need a PolyNeural.ai API key (format: kg_xxxxxxxx)
3. Node.js 16+: Required for running the extension

Installation

Option 1: As MCPB Bundle (Recommended)

1. Download the latest polyneural-mcpb.mcpb file from the releases page
2. Install using your MCPB-compatible application:
   • Claude Desktop: Double-click the .mcpb file to install
   • Other MCPB clients: Follow your client's installation process
3. Configure your PolyNeural.ai API key in the bundle settings UI

Option 2: Manual Installation

1. Install dependencies:

npm install

1. For Claude Desktop: Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "@polyneural/mcpb": {
      "command": "node",
      "args": ["/path/to/mcpb/server/index.js"],
      "env": {
        "API_URL": "https://polyneural.ai",
        "API_KEY": "kg_your_api_key_here",
        "DEBUG": "false",
        "TIMEOUT": "30"
      }
    }
  }
}

Configuration

The MCPB bundle supports these user configuration options (set via MCPB client UI or environment variables):

Available Tools

Once installed, Claude will have access to these PolyNeural.ai knowledge graph tools:

Core Operations

• create_entities - Store new knowledge entities
• create_relations - Create relationships between entities
• search_nodes - Search the knowledge graph
• open_nodes - Retrieve specific entities
• read_graph - Get the complete graph structure

Advanced Operations

• add_observations - Add details to existing entities
• delete_entities - Remove entities
• delete_relations - Remove relationships
• delete_observations - Remove specific observations
• get_entities_by_identifiers - Bulk entity retrieval
• get_entity_relationships - Get entity connections
• get_entities_by_date_range - Time-based queries
• get_recent_changes - Recent activity tracking
• get_trending_entities - Popular entities
• get_frecency_entities - Frequency + recency ranking
• search_with_frecency - Smart search with ranking

Testing

Manual Testing

# Set environment variables and run
API_URL=https://polyneural.ai API_KEY=kg_your_key DEBUG=true npm start

Create MCPB Bundle

# Install MCPB CLI (if not already installed)
npm install -g @anthropic-ai/mcpb

# Create the bundle
mcpb pack

Integration Testing

# Run the integration test script
node test-integration.js

Troubleshooting

Common Issues

Bundle won't start:

• Verify the PolyNeural.ai backend is running: curl https://polyneural.ai/health
• Check your API key format (must start with kg_)
• Enable debug mode in the bundle configuration

No tools available in Claude:

• Ensure Claude Desktop is restarted after installing the bundle
• Check the logs for authentication errors
• Verify the bundle was installed correctly

Connection timeouts:

• Increase the TIMEOUT value
• Check network connectivity to the backend
• Verify the backend MCP endpoints are accessible: curl -H "Authorization: Bearer kg_your_key" https://polyneural.ai/mcp/initialize

Debug Mode

Enable debug logging to see detailed communication:

DEBUG=true npm start

This will show:

• HTTP request details and responses
• Authentication headers
• MCP message routing
• Error details

Architecture Details

MCP Client (stdio MCP client)
    ↕ (JSON-RPC over stdio)
@polyneural/mcpb MCP Server (HTTP bridge)
    ↕ (HTTP with Authorization headers)
PolyNeural.ai Backend (HTTP MCP endpoints)
    ↕ (Database operations)
Knowledge Graph Database

The MCPB server handles:

• stdio MCP server implementation
• HTTP requests to backend MCP endpoints
• Authentication header forwarding
• Request/response translation
• Error handling and logging

Development

To modify this MCPB bundle:

1. Server Logic: Edit server/index.js (HTTP bridge implementation)
2. Bundle Configuration: Modify manifest.json for MCPB-specific settings
3. Dependencies: Update package.json as needed
4. Build Bundle: Run mcpb pack to create the .mcpb file

The beauty of this architecture is that all the actual MCP tool logic remains in the PolyNeural.ai backend - this bundle is purely an HTTP bridge.

MCPB Compliance

This bundle follows the MCPB specification:

• ✅ Valid manifest.json with manifest_version: "0.1"
• ✅ Backward Compatible: Also includes dxt_version: "0.1" for current Claude Desktop
• ✅ Proper MCP server implementation using @modelcontextprotocol/sdk
• ✅ User configuration via user_config field
• ✅ Platform compatibility declarations
• ✅ Proper error handling and timeout management
• ✅ Comprehensive tool and capability declarations

Compatibility Note

The manifest includes both manifest_version (MCPB standard) and dxt_version (current Claude Desktop requirement) to ensure compatibility with both current and future versions of MCPB-compatible applications.