steampipe

This MCP server enables natural language querying and analysis over Steampipe data. It exposes a set of Tools that let your AI assistant run SQL queries against Steampipe, list available Steampipe tables, inspect table schemas, and manage Steampipe plugins. Use the best_practices prompt to tune the LLM’s behavior for Steampipe data, and then ask questions like which AWS accounts are visible, or to generate reports on EC2 instances and their attached storage. The included tools steer the model toward read-only SQL operations, ensuring safe access to your cloud data while enabling security, cost, and compliance insights across AWS, Azure, GCP and many services.

Available Tools:

• steampipe_query: Run SQL queries (PostgreSQL syntax) against Steampipe. Provide a string input with your SQL query. Use CTEs for performance and to structure complex analytics.
• steampipe_table_list: List tables, with optional schema and filter parameters to narrow the results.
• steampipe_table_show: Show details for a specific table, including columns and data types. Optionally specify a schema.
• steampipe_plugin_list: List installed Steampipe plugins (data sources like AWS, GCP, Azure).
• steampipe_plugin_show: Show details for a specific plugin installation (version, memory, config).

Prompts and Capabilities:

• best_practices: Guidance for working with Steampipe data, including SQL style, caching, and when to use WITH clauses vs joins.
• Status resource: Check the current Steampipe connection status and the active connection string.

To use locally, ensure Steampipe is running and the MCP server is reachable by your AI assistant. The default connection is to a local Steampipe instance at postgresql://steampipe@localhost:9193/steampipe. For Turbot Pipes, provide a workspace connection string as an argument to the steampipe-mcp tool. After starting, restart your assistant to apply the new MCP server configuration.

Prerequisites:

• Node.js v16 or higher (includes npx)
• For local use: Steampipe installed and running (steampipe service start)
• For Turbot Pipes: A Turbot Pipes workspace and a connection string

Installation steps:

1. Install dependencies and build the MCP server (from project root):

npm install
npm run build

2. Run the MCP server using npx (as documented):

# Example (default local Steampipe connection)
npx -y @turbot/steampipe-mcp

3. If you plan to point to a Turbot Pipes workspace, pass the connection string as an argument:

npx -y @turbot/steampipe-mcp "postgresql://my_name:my_pw@workspace-name.usea1.db.pipes.turbot.com:9193/abc123"

4. Alternatively, in development testing, run directly via Node.js with the built dist/index.js:

node dist/index.js postgresql://steampipe@localhost:9193/steampipe

5. Verify environment variables:

export STEAMPIPE_MCP_LOG_LEVEL=info
export STEAMPIPE_MCP_WORKSPACE_DATABASE=postgresql://steampipe@localhost:9193/steampipe

6. Optional: Validate MCP server with the Inspector:

npx @modelcontextprotocol/inspector dist/index.js

Environment variables:

• STEAMPIPE_MCP_LOG_LEVEL: Sets the logging verbosity (default: info).
• STEAMPIPE_MCP_WORKSPACE_DATABASE: Overrides the default Steampipe connection string when connecting to a Turbot Pipes workspace. Common issues:
• Ensure Steampipe is running locally if using the default connection string.
• When using Turbot Pipes, ensure the workspace connection string is correct and the network allows access.
• If you encounter permission or read-only errors, verify that the Steampipe user has access to the required data sources and that the MCP server is configured to limit operations to read-only queries.
• Restart the AI assistant after updating the MCP configuration.

Tips:

• Use steampipe_query with CTEs for complex analytics to improve readability and performance.
• Prefer steampipe_table_list and steampipe_table_show to explore the data model before building queries.