Getting Started with NeuronMCP

Introduction

NeuronMCP is a Model Context Protocol (MCP) server providing comprehensive tools and resources for MCP-compatible clients to interact with NeuronDB. It includes:

• MCP Protocol Server - Full JSON-RPC 2.0 implementation with stdio, HTTP, and SSE transport
• 100+ Tools - 27 PostgreSQL administration tools + 70+ NeuronDB tools for vector operations, ML, RAG, and dataset loading
• Resource Provider - Schema, models, indexes, config, workers, and stats with real-time subscriptions
• Enterprise Features - Middleware, authentication, caching, metrics, webhooks, and resilience

What you can build: Connect Claude Desktop and other MCP clients to NeuronDB for vector search, ML training, RAG pipelines, and complete database management - all through the MCP protocol.

Choose Your Path

Pick the installation method that best fits your needs:

Docker Quick Start

Complete NeuronMCP setup running in under 5 minutes with Docker Compose. This method includes NeuronMCP server with NeuronDB integration.

# Clone repository (main branch = 3.0.0-devel)
git clone https://github.com/neurondb-ai/neurondb.git
cd neurondb

# For stable 1.0.0 release, checkout REL1_STABLE branch:
# git checkout REL1_STABLE

# Start NeuronMCP with NeuronDB
docker compose up -d neuronmcp neurondb

# Verify services
docker compose ps

# Test MCP server (stdio mode)
echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | ./NeuronMCP/bin/neuronmcp

This starts:

• NeuronMCP (MCP server) with stdio, HTTP, and SSE transport
• NeuronDB (PostgreSQL with extension) on port 5433
• 100+ MCP tools available for clients
• Resource provider for schema, models, and indexes

Continue to Claude Desktop Setup →

Claude Desktop Setup

Connect Claude Desktop to NeuronMCP to access 100+ tools for vector search, ML training, RAG pipelines, and PostgreSQL administration.

# Edit Claude Desktop configuration
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "neurondb": {
      "command": "neuronmcp",
      "args": ["--transport", "stdio"],
      "env": {
        "DB_HOST": "localhost",
        "DB_PORT": "5433",
        "DB_NAME": "neurondb",
        "DB_USER": "postgres",
        "DB_PASSWORD": "postgres"
      }
    }
  }
}

After configuration:

1. Restart Claude Desktop
2. NeuronMCP tools will be available in Claude Desktop
3. Use tools for vector search, ML training, RAG, and database management

See the Claude Desktop Setup Guide for detailed configuration instructions.

Source Build (Advanced)

For production deployments or custom builds, install from source. This requires Go 1.21+, PostgreSQL 16-18 with NeuronDB extension, and build dependencies.

See the Setup Guide for detailed platform-specific instructions and configuration schema setup.

# Install dependencies
sudo apt-get install -y postgresql-17 postgresql-server-dev-17 build-essential golang-go

# Clone repository
git clone https://github.com/neurondb-ai/neurondb.git
cd neurondb

# For stable 1.0.0 release, checkout REL1_STABLE branch:
# git checkout REL1_STABLE

# Build NeuronMCP
cd NeuronMCP
go build -o neuronmcp ./cmd/server

# Run setup script to configure database schema
./scripts/neuronmcp-setup.sh

# Run NeuronMCP server
./neuronmcp --transport stdio

After building NeuronMCP, configure it to connect to your NeuronDB instance and set up the configuration schema. See the Setup Guide for complete instructions.

Next Steps

After installation, use these guides:

• Claude Desktop Setup - Configure Claude Desktop to connect to NeuronMCP
• Tool Catalog - Browse all 100+ available MCP tools
• Features Documentation - Complete feature reference for MCP protocol and tools
• Setup Guide - Complete configuration schema and setup instructions
• NeuronDB Documentation - Vector search, ML inference, and RAG pipelines
• MCP Integration Example - Complete example showing Claude Desktop configuration