Steampipe

What is Steampipe?

A simple MCP server that acts as a bridge between an AI model (such as Claude) and the Steampipe tool, enabling natural language queries to be translated into Steampipe SQL queries and executed against configured plugins.

How to use Steampipe?

Install prerequisites (Python 3.10+, uv, Steampipe with a configured plugin), then run the MCP Inspector with npx -y @modelcontextprotocol/inspector uv --directory . run steampipe_mcp_server.py. After verifying the tool works, add the server configuration to your MCP-supporting LLM (e.g., Claude) and select the run_steampipe_query tool from the LLM interface.

Key features of Steampipe

• Exposes a single run_steampipe_query tool to the AI model
• Translates natural language requests into Steampipe SQL queries
• Works with any Steampipe plugin (e.g., GitHub) configured locally
• Supports the MCP Inspector for testing and debugging
• Requires only standard MCP and Steampipe infrastructure

Use cases of Steampipe

• Asking an AI assistant to query your GitHub repositories using Steampipe
• Running Steampipe queries by describing what data you need in plain English
• Prototyping a proof-of-concept for AI-powered infrastructure observability

FAQ from Steampipe

What prerequisites are needed to run Steampipe?

Python 3.10 or later, the uv package manager with mcp[cli], a working Steampipe installation with at least one plugin configured (e.g., GitHub with credentials in ~/.steampipe/config/github.spc), an MCP-supporting LLM (e.g., Claude), and Node.js with npx.

How can I test if the server is working correctly?

Run the MCP Inspector: npx -y @modelcontextprotocol/inspector uv --directory . run steampipe_mcp_server.py. In the browser UI, connect, go to the Tools tab, select run_steampipe_query, and supply a valid Steampipe query (e.g., select name, fork_count from github_my_repository).

What should I do if the tool is not found in the Inspector?

Check that the @mcp.tool() decorator is properly applied in the Python script.

How do I investigate execution errors?

Look at the "Result" panel in the Inspector and the server logs (stderr) in your terminal. Inspect whether Steampipe ran, if there was a SQL error, timeout, or JSON parsing issue. Additionally, check the Claude logs: tail -f ~/Library/Logs/Claude/mcp.log or tail -f ~/Library/Logs/Claude/mcp-server-steampipe.log.

What is the security risk of using this server?

The AI model blindly executes any SQL query you generate, so there is a possibility to create and execute arbitrary SQL queries via Steampipe using your configured credentials. Use with caution.