Steampipe

You can connect Claude and other AI assistants to your cloud data through a dedicated MCP server that lets you explore and analyze your whole cloud estate with natural language. It provides safe, read-only access to data from AWS, Azure, GCP, and 100+ services, enabling questions, security checks, cost insights, and more across your infrastructure.

How to use

You interact with the Steampipe MCP server by configuring your AI assistant to connect to it. Once connected, you can ask questions about resources, security, compliance, and costs in natural language. Start with simple inquiries to understand what resources you have in specific accounts or regions, then expand to broader analyses that span multiple clouds.

How to install

Prerequisites and setup steps help you get started quickly.

1. Install Node.js v16 or higher (includes npx). Ensure you can run npm and npx from your command line.

2. If you plan to run locally, install and start Steampipe so the MCP server can connect to it.

3. If you plan to use Turbot Pipes, have a Turbot Pipes workspace and connection string ready.

Configuration and usage details

Add Steampipe MCP to your AI assistant’s configuration. The following snippet demonstrates how to wire the MCP server using the standard local setup.

{
  "mcpServers": {
    "steampipe": {
      "command": "npx",
      "args": [
        "-y",
        "@turbot/steampipe-mcp"
      ]
    }
  }
}

Baseline connection defaults

By default, the MCP server connects to your local Steampipe instance at the standard PostgreSQL URL. Ensure you start Steampipe before attempting to connect.

postgresql://steampipe@localhost:9193/steampipe

Using Turbot Pipes workspace

If you are using Turbot Pipes, provide your workspace connection string as part of the MCP command arguments.

{
  "mcpServers": {
    "steampipe": {
      "command": "npx",
      "args": [
        "-y",
        "@turbot/steampipe-mcp",
        "postgresql://my_name:my_pw@workspace-name.usea1.db.pipes.turbot.com:9193/abc123"
      ]
    }
  }
}

Prompts and best practices

Start by running the best_practices prompt included with the MCP server to guide your LLM on how to work with Steampipe data. Then ask questions in natural language, layering complexity as you build queries.

Capabilities and tools

The MCP server provides a set of tools to query and explore data, list tables, and inspect plugins. Use these to discover available data sources and schemas, then construct efficient queries.

Examples of common tasks

Explore your cloud estate: “What AWS accounts can you see?”

List recent S3 buckets: “Show me all S3 buckets that were created in the last week”

Generate an infrastructure report: “List my EC2 instances with their attached EBS volumes”

Security analysis: “Find any IAM users with access keys that haven't been rotated in the last 90 days”

Compliance insights: “Show me all EC2 instances that don't comply with our tagging standards”

Notes for operators

Be explicit about the resources and accounts you want to analyze, and mention regions when needed. Begin with straightforward queries and gradually add conditions. Your natural language is translated into SQL by the system for you.

Available tools

steampipe_query

Query cloud and security data using SQL. Focus on performance by using CTEs and limiting selected columns.

steampipe_table_list

List all available Steampipe tables, with optional schema and filter parameters.

steampipe_table_show

Show detailed information about a specific table, including columns and data types.

steampipe_plugin_list

List all Steampipe plugins installed to access different data sources.

steampipe_plugin_show

Show details for a specific Steampipe plugin installation, including version and configuration.

best_practices

Prompts that teach best practices for working with Steampipe data, including SQL style, performance, and error handling.

status

Represents the current state of the Steampipe connection, including connection string and status.