Cloudflare Remote PostgreSQL Database MCP Server + GitHub OAuth

This is a Model Context Protocol (MCP) server that enables you to chat with your PostgreSQL database, deployable as a remote MCP server with GitHub OAuth through Cloudflare. This is production ready MCP.

Key Features

• 🗄️ Database Integration with Lifespan: Direct PostgreSQL database connection for all MCP tool calls
• 🛠️ Modular, Single Purpose Tools: Following best practices around MCP tools and their descriptions
• 🔐 Role-Based Access: GitHub username-based permissions for database write operations
• 📊 Schema Discovery: Automatic table and column information retrieval
• 🛡️ SQL Injection Protection: Built-in validation and sanitization
• 📈 Monitoring: Optional Sentry integration for production monitoring
• ☁️ Cloud Native: Powered by Cloudflare Workers for global scale

Modular Architecture

This MCP server uses a clean, modular architecture that makes it easy to extend and maintain:

• src/tools/ - Individual tool implementations in separate files
• registerAllTools() - Centralized tool registration system
• Extensible Design - Add new tools by creating files in tools/ and registering them

This architecture allows you to easily add new database operations, external API integrations, or any other MCP tools while keeping the codebase organized and maintainable.

Transport Protocols

This MCP server supports both modern and legacy transport protocols:

• /mcp - Streamable HTTP (recommended): Uses a single endpoint with bidirectional communication, automatic connection upgrades, and better resilience for network interruptions
• /sse - Server-Sent Events (legacy): Uses separate endpoints for requests/responses, maintained for backward compatibility

For new implementations, use the /mcp endpoint as it provides better performance and reliability.

How It Works

The MCP server provides three main tools for database interaction:

1. listTables - Get database schema and table information (all authenticated users)
2. queryDatabase - Execute read-only SQL queries (all authenticated users)
3. executeDatabase - Execute write operations like INSERT/UPDATE/DELETE (privileged users only)

Authentication Flow: Users authenticate via GitHub OAuth → Server validates permissions → Tools become available based on user's GitHub username.

Security Model:

• All authenticated GitHub users can read data
• Only specific GitHub usernames can write/modify data
• SQL injection protection and query validation built-in

Simple Example First

Want to see a basic MCP server before diving into the full database implementation? Check out src/simple-math.ts - a minimal MCP server with a single calculate tool that performs basic math operations (add, subtract, multiply, divide). This example demonstrates the core MCP components: server setup, tool definition with Zod schemas, and dual transport support (/mcp and /sse endpoints). You can run it locally with wrangler dev --config wrangler-simple.jsonc and test at http://localhost:8789/mcp.

Prerequisites

• Node.js installed on your machine
• A Cloudflare account (free tier works)
• A GitHub account for OAuth setup
• A PostgreSQL database (local or hosted)

Getting Started

Step 1: Install Wrangler CLI

Install Wrangler globally to manage your Cloudflare Workers:

npm install -g wrangler

Step 2: Authenticate with Cloudflare

Log in to your Cloudflare account:

wrangler login

This will open a browser window where you can authenticate with your Cloudflare account.

Step 3: Clone and Setup

Clone the repo directly & install dependencies: npm install.

Environment Variables Setup

Before running the MCP server, you need to configure several environment variables for authentication and database access.

Create Environment Variables File

1. Create your .dev.vars file from the example:

   cp .dev.vars.example .dev.vars

1. Configure all required environment variables in .dev.vars:

   # GitHub OAuth (for authentication)
   GITHUB_CLIENT_ID=your_github_client_id
   GITHUB_CLIENT_SECRET=your_github_client_secret
   COOKIE_ENCRYPTION_KEY=your_random_encryption_key

   # Database Connection
   DATABASE_URL=postgresql://username:password@localhost:5432/database_name

   # Optional: Sentry monitoring
   SENTRY_DSN=https://your-sentry-dsn@sentry.io/project-id
   NODE_ENV=development

Getting GitHub OAuth Credentials

1. Create a GitHub OAuth App for local development:

• Go to GitHub Developer Settings
• Click "New OAuth App"
• Application name: MCP Server (Local Development)
• Homepage URL: http://localhost:8792
• Authorization callback URL: http://localhost:8792/callback
• Click "Register application"

1. Copy your credentials:

• Copy the Client ID and paste it as GITHUB_CLIENT_ID in .dev.vars
• Click "Generate a new client secret", copy it, and paste as GITHUB_CLIENT_SECRET in .dev.vars

Generate Encryption Key

Generate a secure random encryption key for cookie encryption:

openssl rand -hex 32

Copy the output and paste it as COOKIE_ENCRYPTION_KEY in .dev.vars.

Database Setup

1. Set up PostgreSQL using a hosted service like:

• Supabase (recommended for beginners)
• Neon
• Or use local PostgreSQL/Supabase

1. Update the DATABASE_URL in .dev.vars with your connection string:

   DATABASE_URL=postgresql://username:password@host:5432/database_name

#### Connection String Examples:

• Local: postgresql://myuser:mypass@localhost:5432/mydb
• Supabase: postgresql://postgres:your-password@db.your-project.supabase.co:5432/postgres

Database Schema Setup

The MCP server works with any PostgreSQL database schema. It will automatically discover:

• All tables in the public schema
• Column names, types, and constraints
• Primary keys and indexes

Testing the Connection: Once you have your database set up, you can test it by asking the MCP server "What tables are available in the database?" and then querying those tables to explore your data.

Local Development & Testing

Run the server locally:

   wrangler dev

This makes the server available at http://localhost:8792

Testing with MCP Inspector

Use the MCP Inspector to test your server:

1. Install and run Inspector:

   npx @modelcontextprotocol/inspector@latest

1. Connect to your local server:

• Preferred: Enter URL: http://localhost:8792/mcp (streamable HTTP transport - newer, more robust)
• Alternative: Enter URL: http://localhost:8792/sse (SSE transport - legacy support)
• Click "Connect"
• Follow the OAuth prompts to authenticate with GitHub
• Once connected, you'll see the available tools

1. Test the tools:

• Use listTables to see your database structure
• Use queryDatabase to run SELECT queries
• Use executeDatabase (if you have write access) for INSERT/UPDATE/DELETE operations

Production Deployment

#### Set up a KV namespace

• Create the KV namespace:

wrangler kv namespace create "OAUTH_KV"

• Update the wrangler.jsonc file with the KV ID (replace )

#### Deploy Deploy the MCP server to make it available on your workers.dev domain

wrangler deploy

Create environment variables in production

Create a new GitHub OAuth App:

• For the Homepage URL, specify https://mcp-github-oauth.<your-subdomain>.workers.dev
• For the Authorization callback URL, specify https://mcp-github-oauth.<your-subdomain>.workers.dev/callback
• Note your Client ID and generate a Client secret.
• Set all required secrets via Wrangler:

wrangler secret put GITHUB_CLIENT_ID
wrangler secret put GITHUB_CLIENT_SECRET
wrangler secret put COOKIE_ENCRYPTION_KEY  # use: openssl rand -hex 32
wrangler secret put DATABASE_URL
wrangler secret put SENTRY_DSN  # optional (more on Sentry setup below)

#### Test

Test the remote server using Inspector:

npx @modelcontextprotocol/inspector@latest

Enter https://mcp-github-oauth.<your-subdomain>.workers.dev/mcp (preferred) or https://mcp-github-oauth.<your-subdomain>.workers.dev/sse (legacy) and hit connect. Once you go through the authentication flow, you'll see the Tools working:

You now have a remote MCP server deployed!

Database Tools & Access Control

Available Tools