redshift-utils-mcp

Redshift Utils MCP Server

Overview

This project implements a Model Context Protocol (MCP) server designed specifically to interact with Amazon Redshift databases.

It bridges the gap between Large Language Models (LLMs) or AI assistants (like those in Claude, Cursor, or custom applications) and your Redshift data warehouse, enabling secure, standardized data access and interaction. This allows users to query data, understand database structure, and monitoring/diagnostic operations using natural language or AI-driven prompts.

This server is for developers, data analysts, or teams looking to integrate LLM capabilities directly with their Amazon Redshift data environment in a structured and secure manner.

Table of Contents

• Redshift Utils MCP Server
  • Overview
  • Table of Contents
  • Features
  • Prerequisites
  • Installation
    • Install from PyPI (Recommended)
    • Install from Source
  • Configuration
  • Usage
    • Connecting with Claude Desktop / Anthropic Console:
    • Connecting with Claude Code CLI:
    • Connecting with Cursor IDE:
    • Available MCP Resources
    • Available MCP Tools
  • TO DO
  • References

Features

• ✨ Secure Redshift Connection (via Data API): Connects to your Amazon Redshift cluster using the AWS Redshift Data API via Boto3, leveraging AWS Secrets Manager for credentials managed securely via environment variables.
• 🔍 Schema Discovery: Exposes MCP resources for listing schemas and tables within a specified schema.
• 📊 Metadata & Statistics: Provides a tool (handle_inspect_table) to gather detailed table metadata, statistics (like size, row counts, skew, stats staleness), and maintenance status.
• 📝 Read-Only Query Execution: Offers a secure MCP tool (handle_execute_ad_hoc_query) to execute arbitrary SELECT queries against the Redshift database, enabling data retrieval based on LLM requests.
• 📈 Query Performance Analysis: Includes a tool (handle_diagnose_query_performance) to retrieve and analyze the execution plan, metrics, and historical data for a specific query ID.
• 🔍 Table Inspection: Provides a tool (handle_inspect_table) to perform a comprehensive inspection of a table, including design, storage, health, and usage.
• 🩺 Cluster Health Check: Offers a tool (handle_check_cluster_health) to perform a basic or full health assessment of the cluster using various diagnostic queries.
• 🔒 Lock Diagnosis: Provides a tool (handle_diagnose_locks) to identify and report on current lock contention and blocking sessions.
• 📊 Workload Monitoring: Includes a tool (handle_monitor_workload) to analyze cluster workload patterns over a time window, covering WLM, top queries, and resource usage.
• 📝 DDL Retrieval: Offers a tool (handle_get_table_definition) to retrieve the SHOW TABLE output (DDL) for a specified table.
• 🛡️ Input Sanitization: Utilizes parameterized queries via the Boto3 Redshift Data API client where applicable to mitigate SQL injection risks.
• 🧩 Standardized MCP Interface: Adheres to the Model Context Protocol specification for seamless integration with compatible clients (e.g., Claude Desktop, Cursor IDE, custom applications).

Prerequisites

Software:

• Python 3.10+
• uv (recommended package manager) or pip

Infrastructure & Access:

• Access to an Amazon Redshift cluster.
• An AWS account with permissions to use the Redshift Data API (redshift-data:*) and access the specified Secrets Manager secret (secretsmanager:GetSecretValue).
• A Redshift user account whose credentials are stored in AWS Secrets Manager. This user needs the necessary permissions within Redshift to perform the actions enabled by this server (e.g., CONNECT to the database, SELECT on target tables, SELECT on relevant system views like pg_class, pg_namespace, svv_all_schemas, svv_tables, `svv_table_info``). Using a role with the principle of least privilege is strongly recommended. See Security Considerations.

Credentials:

Your Redshift connection details are managed via AWS Secrets Manager, and the server connects using the Redshift Data API. You need:

• The Redshift cluster identifier.
• The database name within the cluster.
• The ARN of the AWS Secrets Manager secret containing the database credentials (username and password).
• The AWS region where the cluster and secret reside.
• Optionally, an AWS profile name if not using default credentials/region.

These details will be configured via environment variables as detailed in the Configuration section.

Installation

Install from PyPI (Recommended)

The easiest way to install the Redshift Utils MCP Server is directly from PyPI:

# Using pip
pip install redshift-utils-mcp

# Using uv (recommended)
uv pip install redshift-utils-mcp

Install from Source

Alternatively, you can install from the source repository:

# Clone the repository
git clone https://github.com/vinodismyname/redshift-utils-mcp.git
cd redshift-utils-mcp

# Install using uv (recommended)
uv sync

# Or install using pip
pip install -e .

Configuration

Set Environment Variables:
This server requires the following environment variables to connect to your Redshift cluster via the AWS Data API. You can set these directly in your shell, using a systemd service file, a Docker environment file, or by creating a .env file in the project's root directory (if using a tool like uv or python-dotenv that supports loading from .env).

Example using shell export:

export REDSHIFT_CLUSTER_ID="your-cluster-id"
export REDSHIFT_DATABASE="your_database_name"
export REDSHIFT_SECRET_ARN="arn:aws:secretsmanager:us-east-1:123456789012:secret:your-redshift-secret-XXXXXX"
export AWS_REGION="us-east-1" # Or AWS_DEFAULT_REGION
# export AWS_PROFILE="your-aws-profile-name" # Optional

Example .env file (see .env.example):

# .env file for Redshift MCP Server configuration
# Ensure this file is NOT committed to version control if it contains secrets. Add it to .gitignore.

REDSHIFT_CLUSTER_ID="your-cluster-id"
REDSHIFT_DATABASE="your_database_name"
REDSHIFT_SECRET_ARN="arn:aws:secretsmanager:us-east-1:123456789012:secret:your-redshift-secret-XXXXXX"
AWS_REGION="us-east-1" # Or AWS_DEFAULT_REGION
# AWS_PROFILE="your-aws-profile-name" # Optional

Required Variables Table:

Variable Name	Required	Description	Example Value
REDSHIFT_CLUSTER_ID	Yes	Your Redshift cluster identifier.	my-redshift-cluster
REDSHIFT_DATABASE	Yes	The name of the database to connect to.	mydatabase
REDSHIFT_SECRET_ARN	Yes	AWS Secrets Manager ARN for Redshift credentials.	arn:aws:secretsmanager:us-east-1:123456789012:secret:mysecret-abcdef
AWS_REGION	Yes	AWS region for Data API and Secrets Manager.	us-east-1
AWS_DEFAULT_REGION	No	Alternative to AWS_REGION for specifying the AWS region.	us-west-2
AWS_PROFILE	No	AWS profile name to use from your credentials file (~/.aws/...).	my-redshift-profile

Note: Ensure the AWS credentials used by Boto3 (via environment, profile, or IAM role) have permissions to access the specified REDSHIFT_SECRET_ARN and use the Redshift Data API (redshift-data:*).

Usage

After installation, you can run the server directly from the command line:

# If installed from PyPI
redshift-utils-mcp

# Or using uvx (no installation required)
uvx redshift-utils-mcp

Connecting with Claude Desktop / Anthropic Console:

Add the following configuration block to your mcp.json file:

{
  "mcpServers": {
    "redshift-utils-mcp": {
      "command": "uvx",
      "args": ["redshift-utils-mcp"],
      "env": {
        "REDSHIFT_CLUSTER_ID":"your-cluster-id",
        "REDSHIFT_DATABASE":"your_database_name",
        "REDSHIFT_SECRET_ARN":"arn:aws:secretsmanager:...",
        "AWS_REGION": "us-east-1"
      }
  }
}

Connecting with Claude Code CLI:

Use the Claude CLI to add the server configuration:

claude mcp add redshift-utils-mcp \
  -e REDSHIFT_CLUSTER_ID="your-cluster-id" \
  -e REDSHIFT_DATABASE="your_database_name" \
  -e REDSHIFT_SECRET_ARN="arn:aws:secretsmanager:..." \
  -e AWS_REGION="us-east-1" \
  -- uvx redshift-utils-mcp

Connecting with Cursor IDE:

1. Start the MCP server locally using the instructions in the Usage / Quickstart section.
2. In Cursor, open the Command Palette (Cmd/Ctrl + Shift + P).
3. Type "Connect to MCP Server" or navigate to the MCP settings.
4. Add a new server connection.
5. Choose the stdio transport type.
6. Enter the command and arguments required to start your server (uvx run redshift_utils_mcp). Ensure any necessary environment variables are available to the command being run.
7. Cursor should detect the server and its available tools/resources.

Available MCP Resources

Resource URI Pattern	Description	Example URI
/scripts/{script_path}	Retrieves the raw content of a SQL script file from the server's sql_scripts directory.	/scripts/health/disk_usage.sql
redshift://schemas	Lists all accessible user-defined schemas in the connected database.	redshift://schemas
redshift://wlm/configuration	Retrieves the current Workload Management (WLM) configuration details.	redshift://wlm/configuration
redshift://schema/{schema_name}/tables	Lists all accessible tables and views within the specified {schema_name}.	redshift://schema/public/tables

Replace {script_path} and {schema_name} with the actual values when making requests.
Accessibility of schemas/tables depends on the permissions granted to the Redshift user configured via REDSHIFT_SECRET_ARN.

Available MCP Tools

Tool Name	Description	Key Parameters (Required*)	Example Invocation
handle_check_cluster_health	Performs a health assessment of the Redshift cluster using a set of diagnostic SQL scripts.	level (optional), time_window_days (optional)	use_mcp_tool("redshift-admin", "handle_check_cluster_health", {"level": "full"})
handle_diagnose_locks	Identifies active lock contention and blocking sessions in the cluster.	min_wait_seconds (optional)	use_mcp_tool("redshift-admin", "handle_diagnose_locks", {"min_wait_seconds": 10})
handle_diagnose_query_performance	Analyzes a specific query's execution performance, including plan, metrics, and historical data.	query_id*	use_mcp_tool("redshift-admin", "handle_diagnose_query_performance", {"query_id": 12345})
handle_execute_ad_hoc_query	Executes an arbitrary SQL query provided by the user via Redshift Data API. Designed as an escape hatch.	sql_query*	use_mcp_tool("redshift-admin", "handle_execute_ad_hoc_query", {"sql_query": "SELECT ..."})
handle_get_table_definition	Retrieves the DDL (Data Definition Language) statement (SHOW TABLE) for a specific table.	schema_name, table_name	use_mcp_tool("redshift-admin", "handle_get_table_definition", {"schema_name": "public", ...})
handle_inspect_table	Retrieves detailed information about a specific Redshift table, covering design, storage, health, and usage.	schema_name, table_name	use_mcp_tool("redshift-admin", "handle_inspect_table", {"schema_name": "analytics", ...})
handle_monitor_workload	Analyzes cluster workload patterns over a specified time window using various diagnostic scripts.	time_window_days (optional), top_n_queries (optional)	use_mcp_tool("redshift-admin", "handle_monitor_workload", {"time_window_days": 7})

TO DO

• Improve Prompt Options
• Add support for more credential methods
• Add Support for Redshift Serverless

References

• This project relies heavily on the Model Context Protocol specification.
• Built using the official MCP SDK provided by Model Context Protocol.
• Utilizes the AWS SDK for Python (Boto3) to interact with the Amazon Redshift Data API.
• Many of the diagnostic SQL scripts are adapted from the excellent awslabs/amazon-redshift-utils repository.