Steampipe MCP Server

What is Steampipe MCP Server?

Steampipe MCP Server lets AI assistants run read-only SQL queries and explore schemas of a PostgreSQL database, primarily designed to work with Steampipe’s multi-cloud connection schemas. It is for users who want to query infrastructure data from AWS, GCP, and other clouds through natural‑language prompts in Claude Desktop or other MCP clients.

How to use Steampipe MCP Server?

First, install Steampipe and start the service to obtain the database URL. Then run the server via steampipe-mcp-server --database-url postgresql://... or configure the STEAMPIPE_MCP_DATABASE_URL environment variable. To use with Claude Desktop, add a steampipe entry in the MCP server configuration pointing to uvx steampipe-mcp-server with the URL.

Key features of Steampipe MCP Server

• Runs read‑only SQL queries via the query tool
• Lists all tables in the search path (list_all_tables)
• Lists tables within a specific schema (list_tables_in_schema)
• Retrieves column names and types for a table (get_table_schema)
• Connects to any PostgreSQL database, optimized for Steampipe

Use cases of Steampipe MCP Server

• Querying cloud resource inventories (EC2 instances, GCP buckets) using natural language
• Exploring available tables and schemas without manual SQL dives
• Running ad‑hoc read‑only queries from an AI assistant in Claude Desktop
• Rapidly prototyping infrastructure reporting across multiple Steampipe connections

FAQ from Steampipe MCP Server

What is the recommended way to prompt Claude when using Steampipe MCP Server?

Specify the schema (e.g., aws_all) so Claude limits table discovery to that schema, avoiding excessive results from large cloud‑provider table sets.

What are the prerequisites for using the server?

You need a running Steampipe PostgreSQL service (started via steampipe service start) and its database URL. The URL must use the postgresql:// protocol.

How do I configure the database connection?

Pass the database URL with the --database-url argument or set the STEAMPIPE_MCP_DATABASE_URL environment variable in a .env file; the server loads it automatically.

What tools does the server provide?

Four tools: query (read‑only SQL), list_all_tables, list_tables_in_schema, and get_table_schema.

Is the server read‑only?

Yes – the query tool only permits read‑only SQL operations; write operations are not supported.