🚀 Initial release of the Redshift Admin MCP Server!
This is the first version of the server, designed to bridge the gap between AI assistants (like Claude, Cursor, etc.) and your Amazon Redshift data warehouse using the Model Context Protocol (MCP). It leverages the AWS Redshift Data API for secure and simplified interactions.
✨ Features
- MCP Server Foundation: Built using the
mcpPython library. - Secure Redshift Connection: Connects via AWS Redshift Data API using Boto3, configured via AWS Secrets Manager ARN and standard AWS credentials (environment variables, profiles, IAM roles).
- Command-Line Interface: Basic CLI powered by
Typerfor starting the server and managing configuration overrides (src/redshift_utils_mcp/__main__.py). - Core MCP Tools Implemented:
handle_check_cluster_health: Performs basic/full health checks using diagnostic SQL scripts.handle_diagnose_locks: Identifies lock contention and blocking PIDs.handle_diagnose_query_performance: Analyzes performance metrics, plan, and history for a specific query ID.handle_execute_ad_hoc_query: Allows execution of arbitrary (read-only intended) SQL queries.handle_get_table_definition: Retrieves table DDL usingSHOW TABLE.handle_inspect_table: Gathers comprehensive details about a table (design, storage, health, usage).handle_monitor_workload: Analyzes workload patterns over a time window.
- MCP Resources Implemented:
/scripts/{script_path}: Serves the content of internal SQL scripts.redshift://schemas: Lists user-defined schemas.redshift://wlm/configuration: Retrieves WLM configuration details.redshift://schema/{schema_name}/tables: Lists tables within a specified schema.
🚀 Usage
-
Configure Environment Variables: Set the required variables for connecting via Redshift Data API. See the Configuration section in the
README.mdfor details. Essential variables are:REDSHIFT_CLUSTER_IDREDSHIFT_DATABASEREDSHIFT_SECRET_ARNAWS_REGION(orAWS_DEFAULT_REGION)- (Optional)
AWS_PROFILE
-
Connect using your preferred MCP client (e.g., Claude Desktop, Cursor IDE) by configuring it to launch the server via
stdio. See the MCP Integration section in theREADME.md.
⚠️ Known Issues / Limitations
- Handler Output Format: Most tool handlers currently return raw results (lists of dictionaries from SQL queries) or Python Exception objects directly within the results dictionary when errors occur during script execution. User-facing formatting or summarization is minimal within the handlers themselves.
- Limited Resource Discovery: Only basic schema, table listing, and WLM config resources are implemented. No resources for views, functions, detailed column info, etc.
- Read-Only Focus: Designed primarily for diagnostics and read operations. No built-in tools for
INSERT,UPDATE,DELETE. Thehandle_execute_ad_hoc_querytool could potentially execute write operations if the configured user has sufficient permissions, use with caution. - SQL Script Dependency: Relies on the bundled SQL scripts. Errors in these scripts or changes in Redshift system views they query could affect functionality.
We hope you find this initial release useful! Please report bugs or suggest features via GitHub Issues.