MCP Server: SSH Rails Runner

What is SSH Rails Runner?

An MCP server that enables secure remote execution of Rails console commands over SSH. It provides tools for read-only operations and carefully managed mutations in a deployed Rails environment, and works well with Cursor for data wrangling and analysis.

How to use SSH Rails Runner?

Install via npm install and npm run build, then set the environment variables SSH_HOST, SSH_USER, SSH_PRIVATE_KEY_PATH, and RAILS_WORKING_DIR. Integrate with MCP clients like Claude Desktop by adding the server to its mcpServers configuration, or use directly with Cursor Composer.

Key features of SSH Rails Runner

• Remote Rails console execution over SSH
• Safe read‑only operations
• Dry‑run capability for mutations
• Execution of approved mutation code snippets
• Resource management for code snippets
• Works with Cursor and Claude Desktop

Use cases of SSH Rails Runner

• Run read‑only Rails console commands on a remote server
• Plan and validate database mutations without side effects
• Execute approved mutations after explicit review
• Perform data analysis and wrangling without a complicated Admin UI

FAQ from SSH Rails Runner

What security considerations should I be aware of?

Only use the server with trusted SSH endpoints from your own local machine that is temporarily provided access to the remote environment. Always review all mutations before execution.

What environment variables are required?

You need to set SSH_HOST, SSH_USER, SSH_PRIVATE_KEY_PATH, and RAILS_WORKING_DIR.

How do I use SSH Rails Runner with Cursor?

In Cursor Composer, pull in your Rails model files as context, then use the execute_read_only, dry_run_mutate, and execute_mutate tools to query and modify the database.

What tools does the server provide?

Three tools: run_read_only for read-only operations, dry_run_mutate to plan and validate mutations without executing them, and execute_mutate to execute previously approved code snippets.

Does SSH Rails Runner support mutations?

Yes, but only through a two‑step process: first create a dry‑run snippet, then explicitly approve and execute it with execute_mutate.