MCPFast / Tools / ShellGuard: Read-only SSH access for LLM agents
An MCP server providing LLM agents with secure, read-only SSH access, powering Fawdy.
View on GitHub→ShellGuard is an MCP server designed to provide Large Language Model (LLM) agents with secure, read-only access to SSH environments. This tool is crucial for developers building AI agents that require interaction with systems without the risk of unintended modifications. By offering a controlled and restricted interface, ShellGuard enables agents to gather information, execute diagnostic commands, and perform other read-only operations safely, powering applications like Fawdy.
ShellGuard acts as an intermediary, allowing LLM agents to connect to SSH servers in a strictly read-only mode. This means agents can list directories, view file contents, check process status, and execute commands that do not alter the system's state. The primary function is to prevent agents from making changes, thereby safeguarding the integrity of the target systems. This is achieved through a carefully configured proxy that filters and restricts outgoing commands.
ShellGuard is an essential tool for AI developers and MLOps engineers who are building or deploying LLM agents that need to interact with production or development servers. It is particularly useful for: