A Model Context Protocol (MCP) server implementation for Nile database platform. This server allows LLM applications to interact with Nile platform through a standardized interface.

Features

• Database Management: Create, list, get details, and delete databases
• Credential Management: Create and list database credentials
• Region Management: List available regions for database creation
• SQL Query Support: Execute SQL queries directly on Nile databases
• MCP Protocol Support: Full implementation of the Model Context Protocol
• Type Safety: Written in TypeScript with full type checking
• Error Handling: Comprehensive error handling and user-friendly error messages
• Test Coverage: Comprehensive test suite using Jest
• Environment Management: Automatic loading of environment variables from .env file
• Input Validation: Schema-based input validation using Zod

Installation

Install the stable version:

npm install @niledatabase/nile-mcp-server

For the latest alpha/preview version:

npm install @niledatabase/nile-mcp-server@alpha

This will install @niledatabase/nile-mcp-server in your node_modules folder. For example: node_modules/@niledatabase/nile-mcp-server/dist/

Manual Installation

Clone the repository

git clone https://github.com/yourusername/nile-mcp-server.git cd nile-mcp-server

Install dependencies

npm install

Build the project

npm run build

Starting the Server

There are several ways to start the server:

1. Direct Node Execution:
2. Development Mode (with auto-rebuild):

The server will start and listen for MCP protocol messages. You should see startup logs indicating:

• Environment variables loaded
• Server instance created
• Tools initialized
• Transport connection established

To stop the server, press Ctrl+C.

Verifying the Server is Running

When the server starts successfully, you should see logs similar to:

[info] Starting Nile MCP Server...
[info] Loading environment variables...
[info] Environment variables loaded successfully
[info] Creating server instance...
[info] Tools initialized successfully
[info] Setting up stdio transport...
[info] Server started successfully

If you see these logs, the server is ready to accept commands from Claude Desktop.

Configuration

Create a .env file in the root directory with your Nile credentials:

NILE_API_KEY=your_api_key_here NILE_WORKSPACE_SLUG=your_workspace_slug

To create a Nile API key, log in to your Nile account, click Workspaces in the top-left, select your workspace, and navigate to the Security section in the left menu.

Using with Claude Desktop

Setup

1. Install Claude Desktop if you haven't already
2. Build the project:
3. Open Claude Desktop
4. Go to Settings > MCP Servers
5. Click "Add Server"
6. Add the following configuration:

{ "mcpServers": { "nile-database": { "command": "node", "args": [ "/path/to/your/nile-mcp-server/dist/index.js" ], "env": { "NILE_API_KEY": "your_api_key_here", "NILE_WORKSPACE_SLUG": "your_workspace_slug" } } } }

Replace:

• /path/to/your/nile-mcp-server with the absolute path to your project directory
• your_api_key_here with your Nile API key
• your_workspace_slug with your Nile workspace slug

Using with Cursor

Setup

1. Install Cursor if you haven't already
2. Build the project:
3. Open Cursor
4. Go to Settings (⌘,) > Features > MCP Servers
5. Click "Add New MCP Server"
6. Configure the server:

   • Name: nile-database (or any name you prefer)

   • Command:

     env NILE_API_KEY=your_key NILE_WORKSPACE_SLUG=your_workspace node /absolute/path/to/nile-mcp-server/dist/index.js

     Replace:

     • your_key with your Nile API key
     • your_workspace with your Nile workspace slug
     • /absolute/path/to with the actual path to your project
7. Click "Save"
8. You should see a green indicator showing that the MCP server is connected
9. Restart Cursor for the changes to take effect

Server Modes

The server supports two operational modes:

STDIO Mode (Default)

The default mode uses standard input/output for communication, making it compatible with Claude Desktop and Cursor integrations.

SSE Mode

Server-Sent Events (SSE) mode enables real-time, event-driven communication over HTTP.

To enable SSE mode:

1. Set MCP_SERVER_MODE=sse in your .env file
2. The server will start an HTTP server (default port 3000)
3. Connect to the SSE endpoint: http://localhost:3000/sse
4. Send commands to: http://localhost:3000/messages

Example SSE usage with curl:

In terminal 1 - Listen for events

curl -N http://localhost:3000/sse

In terminal 2 - Send commands

curl -X POST http://localhost:3000/messages \ -H "Content-Type: application/json" \ -d '{ "type": "function", "name": "list-databases", "parameters": {} }'

Example Prompts

After setting up the MCP server in Cursor, you can use natural language to interact with Nile databases. Here are some example prompts:

Database Management

Create a new database named "my_app" in AWS_US_WEST_2 region

List all my databases

Get details for database "my_app"

Delete database "test_db"

Creating Tables

Create a users table in my_app database with columns:
- tenant_id (UUID, references tenants)
- id (INTEGER)
- email (VARCHAR, unique per tenant)
- name (VARCHAR)
- created_at (TIMESTAMP)

Create a products table in my_app database with columns:
- tenant_id (UUID, references tenants)
- id (INTEGER)
- name (VARCHAR)
- price (DECIMAL)
- description (TEXT)
- created_at (TIMESTAMP)

Querying Data

Execute this query on my_app database:
SELECT * FROM users WHERE tenant_id = 'your-tenant-id' LIMIT 5

Run this query on my_app:
INSERT INTO users (tenant_id, id, email, name) 
VALUES ('tenant-id', 1, 'user@example.com', 'John Doe')

Show me all products in my_app database with price > 100

Schema Management

Show me the schema for the users table in my_app database

Add a new column 'status' to the users table in my_app database

Create an index on the email column of the users table in my_app

Available Tools

The server provides the following tools for interacting with Nile databases:

Database Management

1. create-database

   • Creates a new Nile database
   • Parameters:
     • name (string): Name of the database
     • region (string): Either AWS_US_WEST_2 (Oregon) or AWS_EU_CENTRAL_1 (Frankfurt)
   • Returns: Database details including ID, name, region, and status
   • Example: "Create a database named 'my-app' in AWS_US_WEST_2"

2. list-databases

   • Lists all databases in your workspace
   • No parameters required
   • Returns: List of databases with their IDs, names, regions, and status
   • Example: "List all my databases"

3. get-database

   • Gets detailed information about a specific database
   • Parameters:
     • name (string): Name of the database
   • Returns: Detailed database information including API host and DB host
   • Example: "Get details for database 'my-app'"

4. delete-database

   • Deletes a database
   • Parameters:
     • name (string): Name of the database to delete
   • Returns: Confirmation message
   • Example: "Delete database 'my-app'"

Credential Management

1. list-credentials

   • Lists all credentials for a database
   • Parameters:
     • databaseName (string): Name of the database
   • Returns: List of credentials with IDs, usernames, and creation dates
   • Example: "List credentials for database 'my-app'"

2. create-credential

   • Creates new credentials for a database
   • Parameters:
     • databaseName (string): Name of the database
   • Returns: New credential details including username and one-time password
   • Example: "Create new credentials for database 'my-app'"
   • Note: Save the password when it's displayed, as it won't be shown again

Region Management

1. list-regions
   • Lists all available regions for creating databases
   • No parameters required
   • Returns: List of available AWS regions
   • Example: "What regions are available for creating databases?"

SQL Query Execution

1. execute-sql
   • Executes SQL queries on a Nile database
   • Parameters:
     • databaseName (string): Name of the database to query
     • query (string): SQL query to execute
     • connectionString (string, optional): Pre-existing connection string to use for the query
   • Returns: Query results formatted as a markdown table with column headers and row count
   • Features:
     • Automatic credential management (creates new if not specified)
     • Secure SSL connection to database
     • Results formatted as markdown tables
     • Detailed error messages with hints
     • Support for using existing connection strings
   • Example: "Execute SELECT * FROM users LIMIT 5 on database 'my-app'"

Resource Management

1. read-resource

   • Reads schema information for database resources (tables, views, etc.)
   • Parameters:
     • databaseName (string): Name of the database
     • resourceName (string): Name of the resource (table/view)
   • Returns: Detailed schema information including:
     • Column names and types
     • Primary keys and indexes
     • Foreign key relationships
     • Column descriptions and constraints
   • Example: "Show me the schema for the users table in my-app"

2. list-resources

   • Lists all resources (tables, views) in a database
   • Parameters:
     • databaseName (string): Name of the database
   • Returns: List of all resources with their types
   • Example: "List all tables in my-app database"

Tenant Management

1. list-tenants

   • Lists all tenants in a database
   • Parameters:
     • databaseName (string): Name of the database
   • Returns: List of tenants with their IDs and metadata
   • Example: "Show all tenants in my-app database"

2. create-tenant

   • Creates a new tenant in a database
   • Parameters:
     • databaseName (string): Name of the database
     • tenantName (string): Name for the new tenant
   • Returns: New tenant details including ID
   • Example: "Create a tenant named 'acme-corp' in my-app"

3. delete-tenant

   • Deletes tenants in the database
   • Parameters:
     • databaseName (string): Name of the database
     • tenantName (string): Name for the tenant
   • Returns: Success if the tenant is deleted
   • Example: "Delete tenant named 'acme-corp' in my-app"

Example Usage

Here are some example commands you can use in Claude Desktop:

# Database Management
Please create a new database named "my-app" in the AWS_US_WEST_2 region.
Can you list all my databases?
Get the details for database "my-app".
Delete the database named "test-db".

# Connection String Management
Get a connection string for database "my-app".
# Connection string format: postgres://<user>:<password>@<region>.db.thenile.dev:5432/<database>
# Example: postgres://cred-123:password@us-west-2.db.thenile.dev:5432/my-app

# SQL Queries
Execute SELECT * FROM users LIMIT 5 on database "my-app"
Run this query on my-app database: SELECT COUNT(*) FROM orders WHERE status = 'completed'
Using connection string "postgres://user:pass@host:5432/db", execute this query on my-app: SELECT * FROM products WHERE price > 100

Response Format

All tools return responses in a standardized format:

• Success responses include relevant data and confirmation messages
• Error responses include detailed error messages and HTTP status codes
• SQL query results are formatted as markdown tables
• All responses are formatted for easy reading in Claude Desktop

Error Handling

The server handles various error scenarios:

• Invalid API credentials
• Network connectivity issues
• Invalid database names or regions
• Missing required parameters
• Database operation failures
• SQL syntax errors with helpful hints
• Rate limiting and API restrictions

Troubleshooting

1. If Claude says it can't access the tools:

   • Check that the server path in the configuration is correct
   • Ensure the project is built (npm run build)
   • Verify your API key and workspace slug are correct
   • Restart Claude Desktop

2. If database creation fails:

   • Check your API key permissions
   • Ensure the database name is unique in your workspace
   • Verify the region is one of the supported options

3. If credential operations fail:

   • Verify the database exists and is in the READY state
   • Check that your API key has the necessary permissions

Development

Project Structure

nile-mcp-server/
├── src/
│   ├── server.ts      # MCP server implementation
│   ├── tools.ts       # Tool implementations
│   ├── types.ts       # Type definitions
│   ├── logger.ts      # Logging utilities
│   ├── index.ts       # Entry point
│   └── __tests__/     # Test files
│       └── server.test.ts
├── dist/             # Compiled JavaScript
├── logs/            # Log files directory
├── .env             # Environment configuration
├── .gitignore       # Git ignore file
├── package.json     # Project dependencies
└── tsconfig.json    # TypeScript configuration

Key Files

• server.ts: Main server implementation with tool registration and transport handling
• tools.ts: Implementation of all database operations and SQL query execution
• types.ts: TypeScript interfaces for database operations and responses
• logger.ts: Structured logging with daily rotation and debug support
• index.ts: Server startup and environment configuration
• server.test.ts: Comprehensive test suite for all functionality

Development

Install dependencies

npm install

Build the project

npm run build

Start the server in production mode

node dist/index.js

Start the server using npm script

npm start

Start in development mode with auto-rebuild

npm run dev

Run tests

npm test

Development Scripts

The following npm scripts are available:

• npm run build: Compiles TypeScript to JavaScript
• npm start: Starts the server in production mode
• npm run dev: Starts the server in development mode with auto-rebuild
• npm test: Runs the test suite
• npm run lint: Runs ESLint for code quality checking
• npm run clean: Removes build artifacts

Testing

The project includes a comprehensive test suite that covers:

• Tool registration and schema validation
• Database management operations
• Connection string generation
• SQL query execution and error handling
• Response formatting and error cases

Run the tests with:

Logging

The server uses structured logging with the following features:

• Daily rotating log files
• Separate debug logs
• JSON formatted logs with timestamps
• Console output for development
• Log categories: info, error, debug, api, sql, startup

License

MIT License - See LICENSE for details.

Related Links

• Model Context Protocol
• Nile Database
• Claude Desktop
• Cursor