AWS Labs Aurora DSQL MCP Server

An AWS Labs Model Context Protocol (MCP) server for Aurora DSQL and corresponding AI rules that can be used for additional model steering while developing.

Features

• Converting human-readable questions and commands into structured Postgres-compatible SQL queries and executing them against the configured Aurora DSQL database.
• Read-only by default, transactions enabled with --allow-writes
• Connection reuse between requests for improved performance
• Built-in access to Aurora DSQL documentation, search, and best practice recommendations

Available Tools

Database Operations

[IMPORTANT] The MCP Server requires a valid configuration for --cluster_endpoint, --database_user, and --region to enable database operations.

• readonly_query - Execute read-only SQL queries against your DSQL cluster
• transact - Execute SQL statements in a transaction
  • In read-only mode: Supports read operations with transactional consistency
  • With --allow-writes: Supports all write operations too
• get_schema - Retrieve table schema information

Documentation and Recommendations

• dsql_search_documentation - Search Aurora DSQL documentation
  • Parameters: search_phrase (required), limit (optional)
• dsql_read_documentation - Read specific DSQL documentation pages
  • Parameters: url (required), start_index (optional), max_length (optional)
• dsql_recommend - Get recommendations for DSQL best practices
  • Parameters: url (required)

SQL Validation

• dsql_lint - Validate SQL for Aurora DSQL compatibility and optionally auto-fix issues
  • Parameters: sql (required), fix (optional, default false)
  • Returns diagnostics with rule violations, suggestions, and optionally a fixed SQL string

Prerequisites

1. An AWS account with an Aurora DSQL Cluster
2. This MCP server can only be run locally on the same host as your LLM client.
3. Set up AWS credentials with access to AWS services
   • You need an AWS account with appropriate permissions
   • Configure AWS credentials with aws configure or environment variables

Installation

Kiro	Cursor	VS Code

Using uv

1. Install uv from Astral or the GitHub README
2. Install Python using uv python install 3.10

Configure the MCP server in your MCP client configuration (e.g., for Kiro, edit ~/.kiro/settings/mcp.json):

{
  "mcpServers": {
    "awslabs.aurora-dsql-mcp-server": {
      "command": "uvx",
      "args": [
        "awslabs.aurora-dsql-mcp-server@latest",
        "--cluster_endpoint",
        "[your dsql cluster endpoint]",
        "--region",
        "[your dsql cluster region, e.g. us-east-1]",
        "--database_user",
        "[your dsql username]",
        "--profile",
        "default"
      ],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Windows Installation

For Windows users, the MCP server configuration format is slightly different:

{
  "mcpServers": {
    "awslabs.aurora-dsql-mcp-server": {
      "disabled": false,
      "timeout": 60,
      "type": "stdio",
      "command": "uv",
      "args": [
        "tool",
        "run",
        "--from",
        "awslabs.aurora-dsql-mcp-server@latest",
        "awslabs.aurora-dsql-mcp-server.exe"
      ],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR",
        "AWS_PROFILE": "your-aws-profile",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Using Docker

1. 'git clone https://github.com/awslabs/mcp.git'
2. Go to sub-directory 'src/aurora-dsql-mcp-server/'
3. Run 'docker build -t awslabs/aurora-dsql-mcp-server:latest .'
4. Create a env file with temporary credentials:

Either manually:

# fictitious `.env` file with AWS temporary credentials
AWS_ACCESS_KEY_ID=<from the profile you set up>
AWS_SECRET_ACCESS_KEY=<from the profile you set up>
AWS_SESSION_TOKEN=<from the profile you set up>

Or using aws configure:

aws configure export-credentials --profile your-profile-name --format env > temp_aws_credentials.env | sed 's/^export //' > temp_aws_credentials.env

{
  "mcpServers": {
    "awslabs.aurora-dsql-mcp-server": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "--env-file",
        "/full/path/to/file/above/.env",
        "awslabs/aurora-dsql-mcp-server:latest",
        "--cluster_endpoint",
        "[your data]",
        "--database_user",
        "[your data]",
        "--region",
        "[your data]"
      ]
    }
  }
}

Server Configuration options

--allow-writes

By default, the DSQL MCP server operates in read-only mode. In this mode:

• readonly_query: Executes single read-only queries
• transact: Executes read-only transactions with point-in-time consistency
  • Useful for multiple queries that need to see data at the same point in time
  • All statements are validated to ensure they are read-only operations
  • Write operations (INSERT, UPDATE, DELETE, CREATE, DROP, ALTER, etc.) are rejected

To enable write operations, pass the --allow-writes parameter. In read-write mode:

• readonly_query: Same behavior (read-only queries)
• transact: Supports all DDL and DML operations (CREATE, INSERT, UPDATE, DELETE, etc.)

We recommend using least-privilege access when connecting to DSQL. For example, users should use a role that is read-only when possible. The read-only mode provides best-effort client-side validation to reject mutations.

--cluster_endpoint

This is mandatory parameter to specify the cluster to connect to. This should be the full endpoint of your cluster, e.g., 01abc2ldefg3hijklmnopqurstu.dsql.us-east-1.on.aws

--database_user

This is a mandatory parameter to specify the user to connect as. For example admin, or my_user. Note that the AWS credentials you are using must have permission to login as that user. For more information on setting up and using database roles in DSQL, see Using database roles with IAM roles.

--profile

You can specify the aws profile to use for your credentials. Note that this is not supported for docker installation.

Using the AWS_PROFILE environment variable in your MCP configuration is also supported:

"env": {
  "AWS_PROFILE": "your-aws-profile"
}

If neither is provided, the MCP server defaults to using the "default" profile in your AWS configuration file.

--region

This is a mandatory parameter to specify the region of your DSQL database.

--knowledge-server

Optional parameter to specify the remote MCP server endpoint for DSQL knowledge tools (documentation search, reading, and recommendations). By default it is pre-configured.

Example:

--knowledge-server https://custom-knowledge-server.example.com

Note: For security, only use trusted knowledge server endpoints. The server should be an HTTPS endpoint.

--knowledge-timeout

Optional parameter to specify the timeout in seconds for requests to the knowledge server.

Default: 30.0

Example:

--knowledge-timeout 60.0

Increase this value if you experience timeouts when accessing documentation on slow networks.

Development and Testing

Running Tests

This project includes comprehensive tests to validate the readonly enforcement mechanisms. To run the tests:

# Install dependencies and run tests
uv run pytest tests/test_readonly_enforcement.py -v

# Run all tests
uv run pytest -v

# Run tests with coverage
uv run pytest --cov=awslabs.aurora_dsql_mcp_server tests/ -v

Local Docker Testing

To test the MCP server locally using Docker:

1. Build the Docker image:

   cd src/aurora-dsql-mcp-server
   docker build -t awslabs/aurora-dsql-mcp-server:latest .
   

2. Create AWS credentials file:

   Option A - Manual creation:

   # Create .env file with your AWS credentials
   cat > .env << EOF
   AWS_ACCESS_KEY_ID=your_access_key_here
   AWS_SECRET_ACCESS_KEY=your_secret_key_here
   AWS_SESSION_TOKEN=your_session_token_here
   EOF
   

   Option B - Export from AWS CLI:

   aws configure export-credentials --profile your-profile-name --format env > temp_aws_credentials.env
   sed 's/^export //' temp_aws_credentials.env > .env
   rm temp_aws_credentials.env
   

3. Test the container directly:

   docker run -i --rm \
     --env-file .env \
     awslabs/aurora-dsql-mcp-server:latest \
     --cluster_endpoint "your-dsql-cluster-endpoint" \
     --database_user "your-username" \
     --region "us-east-1"
   

4. Test with write operations enabled:

   docker run -i --rm \
     --env-file .env \
     awslabs/aurora-dsql-mcp-server:latest \
     --cluster_endpoint "your-dsql-cluster-endpoint" \
     --database_user "your-username" \
     --region "us-east-1" \
     --allow-writes
   

Note: Replace the placeholder values with your actual DSQL cluster endpoint, username, and region.

AI Rules

The standalone DSQL Agent Skill in this repository is deprecated. Its directories now provide compatibility redirects only. Use the canonical dsql skill in the databases-on-aws plugin for current DSQL guidance.

The Kiro Power remains available as described below.

Kiro Power

To setup the Kiro power:

1. Install directly from the Kiro Powers Registry
2. Once redirected to the Power in the IDE either:
   1. Select the Try Power button. Suggested for people who want:
      • The AI to guide MCP server setup
      • An interactive onboarding experience with DSQL to create a new cluster
   2. Open a new Kiro chat and ask anything related to DSQL
      • Optionally update the MCP Config: Add your existing cluster details and test the MCP server connection so the MCP server can be used out of the box with the power.
      • The Kiro agent will automatically activate the power if it identifies the power as valuable for completing the user's task.

Canonical DSQL Agent Skill

For Claude Code, use the verified plugin commands:

/plugin marketplace add awslabs/agent-plugins
/plugin install databases-on-aws@agent-plugins-for-aws

For Codex, Cursor, and other supported clients, follow the current awslabs/agent-plugins installation instructions. Do not remove an existing standalone skill checkout or symlink without the user's explicit approval.