DBHub

DBHub is a universal database gateway implementing the Model Context Protocol (MCP) server interface. This gateway allows MCP-compatible clients to connect to and explore different databases.

 +------------------+    +--------------+    +------------------+
 |                  |    |              |    |                  |
 |                  |    |              |    |                  |
 |  Claude Desktop  +--->+              +--->+    PostgreSQL    |
 |                  |    |              |    |                  |
 |  Claude Code     +--->+              +--->+    SQL Server    |
 |                  |    |              |    |                  |
 |  Cursor          +--->+    DBHub     +--->+    SQLite        |
 |                  |    |              |    |                  |
 |  VS Code         +--->+              +--->+    MySQL         |
 |                  |    |              |    |                  |
 |  Other Clients   +--->+              +--->+    MariaDB       |
 |                  |    |              |    |                  |
 |                  |    |              |    |                  |
 +------------------+    +--------------+    +------------------+
      MCP Clients           MCP Server             Databases

Supported Matrix

Database Resources

Database Tools

Prompt Capabilities

Installation

Docker

# PostgreSQL example
docker run --rm --init \
   --name dbhub \
   --publish 8080:8080 \
   bytebase/dbhub \
   --transport http \
   --port 8080 \
   --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

# Demo mode with sqlite sample employee database
docker run --rm --init \
   --name dbhub \
   --publish 8080:8080 \
   bytebase/dbhub \
   --transport http \
   --port 8080 \
   --demo

Docker Compose Setup:

If you're using Docker Compose for development, add DBHub to your docker-compose.yml:

dbhub:
  image: bytebase/dbhub:latest
  container_name: dbhub
  ports:
    - "8080:8080"
  environment:
    - DBHUB_LOG_LEVEL=info
  command:
    - --transport
    - http
    - --port
    - "8080"
    - --dsn
    - "postgres://user:password@database:5432/dbname"
  depends_on:
    - database

NPM

# PostgreSQL example
npx @bytebase/dbhub --transport http --port 8080 --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

# Demo mode with sqlite sample employee database
npx @bytebase/dbhub --transport http --port 8080 --demo

# Demo mode with sample employee database
npx @bytebase/dbhub --transport http --port 8080 --demo

Note: The demo mode includes a bundled SQLite sample "employee" database with tables for employees, departments, salaries, and more.

Claude Desktop

• Claude Desktop only supports stdio transport https://github.com/orgs/modelcontextprotocol/discussions/16

// claude_desktop_config.json
{
  "mcpServers": {
    "dbhub-postgres-docker": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "bytebase/dbhub",
        "--transport",
        "stdio",
        "--dsn",
        // Use host.docker.internal as the host if connecting to the local db
        "postgres://user:password@host.docker.internal:5432/dbname?sslmode=disable"
      ]
    },
    "dbhub-postgres-npx": {
      "command": "npx",
      "args": [
        "-y",
        "@bytebase/dbhub",
        "--transport",
        "stdio",
        "--dsn",
        "postgres://user:password@localhost:5432/dbname?sslmode=disable"
      ]
    },
    "dbhub-demo": {
      "command": "npx",
      "args": ["-y", "@bytebase/dbhub", "--transport", "stdio", "--demo"]
    }
  }
}

Claude Code

Check https://docs.anthropic.com/en/docs/claude-code/mcp

Cursor

cursor://anysphere.cursor-deeplink/mcp/install?name=dbhub&config=eyJjb21tYW5kIjoibnB4IEBieXRlYmFzZS9kYmh1YiIsImVudiI6eyJUUkFOU1BPUlQiOiJzdGRpbyIsIkRTTiI6InBvc3RncmVzOi8vdXNlcjpwYXNzd29yZEBsb2NhbGhvc3Q6NTQzMi9kYm5hbWU%2Fc3NsbW9kZT1kaXNhYmxlIiwiUkVBRE9OTFkiOiJ0cnVlIn19

• Cursor supports both stdio and http.

• Follow Cursor MCP guide and make sure to use Agent mode.

VSCode + Copilot

Check https://code.visualstudio.com/docs/copilot/customization/mcp-servers

VSCode with GitHub Copilot can connect to DBHub via both stdio and http transports. This enables AI agents to interact with your development database through a secure interface.

• VSCode supports both stdio and http transports

• Configure MCP server in .vscode/mcp.json:

Stdio Transport:

{
  "servers": {
    "dbhub": {
      "command": "npx",
      "args": [
        "-y",
        "@bytebase/dbhub",
        "--transport",
        "stdio",
        "--dsn",
        "postgres://user:password@localhost:5432/dbname"
      ]
    }
  },
  "inputs": []
}

HTTP Transport:

{
  "servers": {
    "dbhub": {
      "url": "http://localhost:8080/mcp",
      "type": "http"
    }
  },
  "inputs": []
}

Copilot Instructions:

You can provide Copilot with context by creating .github/copilot-instructions.md:

## Database Access

This project provides an MCP server (DBHub) for secure SQL access to the development database.

AI agents can execute SQL queries. In read-only mode (recommended for production):

- `SELECT * FROM users LIMIT 5;`
- `SHOW TABLES;`
- `DESCRIBE table_name;`

In read-write mode (development/testing):

- `INSERT INTO users (name, email) VALUES ('John', 'john@example.com');`
- `UPDATE users SET status = 'active' WHERE id = 1;`
- `CREATE TABLE test_table (id INT PRIMARY KEY);`

Use `--readonly` flag to restrict to read-only operations for safety.

Usage

Read-only Mode

You can run DBHub in read-only mode, which restricts SQL query execution to read-only operations:

# Enable read-only mode
npx @bytebase/dbhub --readonly --dsn "postgres://user:password@localhost:5432/dbname"

In read-only mode, only readonly SQL operations are allowed.

This provides an additional layer of security when connecting to production databases.

Suffix Tool Names with ID

You can suffix tool names with a custom ID using the --id flag or ID environment variable. This is useful when running multiple DBHub instances (e.g., in Cursor) to allow the client to route queries to the correct database.

Example configuration for multiple databases in Cursor:

{
  "mcpServers": {
    "dbhub-prod": {
      "command": "npx",
      "args": ["-y", "@bytebase/dbhub"],
      "env": {
        "ID": "prod",
        "DSN": "postgres://user:password@prod-host:5432/dbname"
      }
    },
    "dbhub-staging": {
      "command": "npx",
      "args": ["-y", "@bytebase/dbhub"],
      "env": {
        "ID": "staging",
        "DSN": "mysql://user:password@staging-host:3306/dbname"
      }
    }
  }
}

With this configuration:

• Production database tools: execute_sql_prod

• Staging database tools: execute_sql_staging

Multi-Database Configuration (TOML)

New in v0.11.6+: DBHub now supports TOML-based configuration files for managing multiple database connections within a single instance. This is the recommended approach for projects that need to query across multiple databases.

Quick Start

1. Create a dbhub.toml file in your project directory:

[[sources]]
id = "prod_pg"
dsn = "postgres://user:password@localhost:5432/production"
readonly = false
max_rows = 1000

[[sources]]
id = "staging_mysql"
type = "mysql"
host = "localhost"
port = 3306
database = "staging"
user = "root"
password = "secret"
max_rows = 500

2. Run DBHub:

# Automatically loads ./dbhub.toml
npx @bytebase/dbhub --transport http --port 8080

# Or specify custom config location
npx @bytebase/dbhub --config=/path/to/config.toml --transport http --port 8080

Configuration Structure

Each database source is defined with [[sources]] and requires a unique id field:

Using DSN (recommended):

[[sources]]
id = "my_database"
dsn = "postgres://user:pass@localhost:5432/dbname"
readonly = false
max_rows = 1000

Using individual parameters:

[[sources]]
id = "my_database"
type = "mysql"            # postgres, mysql, mariadb, sqlserver, sqlite
host = "localhost"
port = 3306
database = "mydb"
user = "root"
password = "secret"
readonly = false
max_rows = 1000

With SSH tunnel:

[[sources]]
id = "remote_db"
dsn = "postgres://user:pass@10.0.0.5:5432/dbname"
ssh_host = "bastion.example.com"
ssh_port = 22
ssh_user = "ubuntu"
ssh_key = "~/.ssh/id_rsa"
# ssh_password = "password"  # Alternative to ssh_key
# ssh_passphrase = "key_pass"  # If key is encrypted

Using Multiple Databases in MCP Tools

When using multi-database configuration, you can specify which database to query using the source_id parameter:

// Execute SQL on specific database
execute_sql({
  sql: "SELECT * FROM users LIMIT 10",
  source_id: "prod_pg"  // Uses the source with id="prod_pg"
});

// Execute on default database (first in config)
execute_sql({
  sql: "SELECT * FROM orders LIMIT 10"
  // No source_id specified - uses first [[sources]] entry
});

Configuration Priority

DBHub follows this priority order for configuration:

1. TOML config file (e.g., --config or ./dbhub.toml) OR CLI arguments (e.g., --dsn)
   (These are mutually exclusive: if --config is provided, TOML is used; otherwise, --dsn is used.)

2. Environment variables (e.g., DSN)

3. .env files (e.g., .env or .env.local)

Note: --config (TOML) and --dsn (CLI) are alternative primary configuration methods, both taking precedence over environment variables and .env files. Users cannot combine --dsn with TOML configuration.

Example: Complete Multi-Database Setup

See dbhub.toml.example in the repository for a comprehensive example with all supported options and detailed comments.

Key Features:

• Default database: The first [[sources]] entry is the default

• Per-source settings: Each source can have its own readonly and max_rows settings

• Path expansion: Paths starting with ~/ are expanded to your home directory

• Security: Passwords in DSN strings are automatically redacted in logs

Backward Compatibility

Single-database configurations using CLI arguments, environment variables, or .env files continue to work exactly as before. TOML configuration is optional and only needed for multi-database scenarios.

Row Limiting

You can limit the number of rows returned from SELECT queries using the --max-rows parameter. This helps prevent accidentally retrieving too much data from large tables:

# Limit SELECT queries to return at most 1000 rows
npx @bytebase/dbhub --dsn "postgres://user:password@localhost:5432/dbname" --max-rows 1000

• Row limiting is only applied to SELECT statements, not INSERT/UPDATE/DELETE

• If your query already has a LIMIT or TOP clause, DBHub uses the smaller value

SSL Connections

You can specify the SSL mode using the sslmode parameter in your DSN string:

SSL Mode Options:

• sslmode=disable: All SSL/TLS encryption is turned off. Data is transmitted in plaintext.

• sslmode=require: Connection is encrypted, but the server's certificate is not verified. This provides protection against packet sniffing but not against man-in-the-middle attacks. You may use this for trusted self-signed CA.

Without specifying sslmode, most databases default to certificate verification, which provides the highest level of security.

Example usage:

# Disable SSL
postgres://user:password@localhost:5432/dbname?sslmode=disable

# Require SSL without certificate verification
postgres://user:password@localhost:5432/dbname?sslmode=require

# Standard SSL with certificate verification (default)
postgres://user:password@localhost:5432/dbname

SSH Tunnel Support

DBHub supports connecting to databases through SSH tunnels, enabling secure access to databases in private networks or behind firewalls.

Using SSH Config File (Recommended)

DBHub can read SSH connection settings from your ~/.ssh/config file. Simply use the host alias from your SSH config:

# If you have this in ~/.ssh/config:
# Host mybastion
#   HostName bastion.example.com
#   User ubuntu
#   IdentityFile ~/.ssh/id_rsa

npx @bytebase/dbhub \
  --dsn "postgres://dbuser:dbpass@database.internal:5432/mydb" \
  --ssh-host mybastion

DBHub will automatically use the settings from your SSH config, including hostname, user, port, and identity file. If no identity file is specified in the config, DBHub will try common default locations (~/.ssh/id_rsa, ~/.ssh/id_ed25519, etc.).

SSH with Password Authentication

npx @bytebase/dbhub \
  --dsn "postgres://dbuser:dbpass@database.internal:5432/mydb" \
  --ssh-host bastion.example.com \
  --ssh-user ubuntu \
  --ssh-password mypassword

SSH with Private Key Authentication

npx @bytebase/dbhub \
  --dsn "postgres://dbuser:dbpass@database.internal:5432/mydb" \
  --ssh-host bastion.example.com \
  --ssh-user ubuntu \
  --ssh-key ~/.ssh/id_rsa

SSH with Private Key and Passphrase

npx @bytebase/dbhub \
  --dsn "postgres://dbuser:dbpass@database.internal:5432/mydb" \
  --ssh-host bastion.example.com \
  --ssh-port 2222 \
  --ssh-user ubuntu \
  --ssh-key ~/.ssh/id_rsa \
  --ssh-passphrase mykeypassphrase

Using Environment Variables

export SSH_HOST=bastion.example.com
export SSH_USER=ubuntu
export SSH_KEY=~/.ssh/id_rsa
npx @bytebase/dbhub --dsn "postgres://dbuser:dbpass@database.internal:5432/mydb"

Note: When using SSH tunnels, the database host in your DSN should be the hostname/IP as seen from the SSH server (bastion host), not from your local machine.

Configure your database connection

You can use DBHub in demo mode with a sample employee database for testing:

npx @bytebase/dbhub  --demo

For real databases, you can configure the database connection in two ways:

Method 1: Database Source Name (DSN)

• Command line argument (highest priority):

  npx @bytebase/dbhub  --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

• Environment variable (second priority):

  export DSN="postgres://user:password@localhost:5432/dbname?sslmode=disable"
  npx @bytebase/dbhub

• Environment file (third priority):

  • For development: Create .env.local with your DSN

  • For production: Create .env with your DSN

  DSN=postgres://user:password@localhost:5432/dbname?sslmode=disable

Method 2: Individual Database Parameters

If your password contains special characters that would break URL parsing, use individual environment variables instead:

• Environment variables:

  export DB_TYPE=postgres
  export DB_HOST=localhost
  export DB_PORT=5432
  export DB_USER=myuser
  export DB_PASSWORD='my@complex:password/with#special&chars'
  export DB_NAME=mydatabase
  npx @bytebase/dbhub

• Environment file:

  DB_TYPE=postgres
  DB_HOST=localhost
  DB_PORT=5432
  DB_USER=myuser
  DB_PASSWORD=my@complex:password/with#special&chars
  DB_NAME=mydatabase

Supported DB_TYPE values: postgres, mysql, mariadb, sqlserver, sqlite

Default ports (when DB_PORT is omitted):

• PostgreSQL: 5432

• MySQL/MariaDB: 3306

• SQL Server: 1433

For SQLite: Only DB_TYPE=sqlite and DB_NAME=/path/to/database.db are required.

DBHub supports the following database connection string formats:

SQL Server

Extra query parameters:

authentication

• authentication=azure-active-directory-access-token. Only applicable when running from Azure. See DefaultAzureCredential.

Transport

• stdio (default) - for direct integration with tools like Claude Desktop:

  npx @bytebase/dbhub --transport stdio --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

• http - for browser and network clients:

  npx @bytebase/dbhub --transport http --port 5678 --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

Command line options

The demo mode uses an in-memory SQLite database loaded with the sample employee database that includes tables for employees, departments, titles, salaries, department employees, and department managers. The sample database includes SQL scripts for table creation, data loading, and testing.

Development

1. Install dependencies:

   pnpm install

2. Run in development mode:

   pnpm dev

3. Build for production:

   pnpm build
   pnpm start --transport stdio --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"

Testing

The project uses Vitest for comprehensive unit and integration testing:

• Run all tests: pnpm test

• Run tests in watch mode: pnpm test:watch

• Run integration tests: pnpm test:integration

Integration Tests

DBHub includes comprehensive integration tests for all supported database connectors using Testcontainers. These tests run against real database instances in Docker containers, ensuring full compatibility and feature coverage.

Prerequisites

• Docker: Ensure Docker is installed and running on your machine

• Docker Resources: Allocate sufficient memory (recommended: 4GB+) for multiple database containers

• Network Access: Ability to pull Docker images from registries

Running Integration Tests

Note: This command runs all integration tests in parallel, which may take 5-15 minutes depending on your system resources and network speed.

# Run all database integration tests
pnpm test:integration

# Run only PostgreSQL integration tests
pnpm test src/connectors/__tests__/postgres.integration.test.ts
# Run only MySQL integration tests
pnpm test src/connectors/__tests__/mysql.integration.test.ts
# Run only MariaDB integration tests
pnpm test src/connectors/__tests__/mariadb.integration.test.ts
# Run only SQL Server integration tests
pnpm test src/connectors/__tests__/sqlserver.integration.test.ts
# Run only SQLite integration tests
pnpm test src/connectors/__tests__/sqlite.integration.test.ts
# Run JSON RPC integration tests
pnpm test src/__tests__/json-rpc-integration.test.ts

All integration tests follow these patterns:

1. Container Lifecycle: Start database container → Connect → Setup test data → Run tests → Cleanup

2. Shared Test Utilities: Common test patterns implemented in IntegrationTestBase class

3. Database-Specific Features: Each database includes tests for unique features and capabilities

4. Error Handling: Comprehensive testing of connection errors, invalid SQL, and edge cases

Troubleshooting Integration Tests

Container Startup Issues:

# Check Docker is running
docker ps

# Check available memory
docker system df

# Pull images manually if needed
docker pull postgres:15-alpine
docker pull mysql:8.0
docker pull mariadb:10.11
docker pull mcr.microsoft.com/mssql/server:2019-latest

SQL Server Timeout Issues:

• SQL Server containers require significant startup time (3-5 minutes)

• Ensure Docker has sufficient memory allocated (4GB+ recommended)

• Consider running SQL Server tests separately if experiencing timeouts

Network/Resource Issues:

# Run tests with verbose output
pnpm test:integration --reporter=verbose

# Run single database test to isolate issues
pnpm test:integration -- --testNamePattern="PostgreSQL"

# Check Docker container logs if tests fail
docker logs <container_id>

Debug with MCP Inspector

stdio

# PostgreSQL example
TRANSPORT=stdio DSN="postgres://user:password@localhost:5432/dbname?sslmode=disable" npx @modelcontextprotocol/inspector node /path/to/dbhub/dist/index.js

HTTP

# Start DBHub with HTTP transport
pnpm dev --transport=http --port=8080

# Start the MCP Inspector in another terminal
npx @modelcontextprotocol/inspector

Connect to the DBHub server /mcp endpoint

Contributors