mcpsshclient

What is mcpsshclient?

mcpsshclient is an MCP server that provides an SSH client with a configurable agentic security agent capable of detecting and preventing "unsafe" commands from executing through an SSH connection. It is designed for developers who need secure remote command execution with automated safety checks.

How to use mcpsshclient?

Clone the repository, run npm install and npm run build, then configure your MCP client (e.g., Claude Desktop, Cursor AI, VSCode Insider) to use the built index.js file with the node command. To enable the security agent, set ENABLE_SECAGENT to true and define a SECURITY_POLICY in secagentconfig.json. Prerequisites include installing Ollama, running ollama serve, pulling the llama2 model, and running it.

Key features of mcpsshclient

• Configurable agentic security agent for command detection
• Prevents "unsafe" commands from executing via SSH
• Integrates with Claude Desktop, Cursor AI, and VSCode Insider
• Local AI-based safety enforcement using Ollama and llama2
• JSON-based security policy configuration

Use cases of mcpsshclient

• Secure remote server administration with command guardrails
• Automating SSH tasks while preventing dangerous operations
• Integrating AI-powered safety checks into MCP-enabled client workflows

FAQ from mcpsshclient

What dependencies or runtime requirements does mcpsshclient have?

The server requires Node.js and npm to build and run. For the agentic security features, Ollama must be installed and running locally, with the llama2 model pulled and available.

Where does configuration for the security agent live?

The security agent is configured in a secagentconfig.json file where you set ENABLE_SECAGENT to true and define your SECURITY_POLICY.

Which MCP clients have been tested with mcpsshclient?

It has been tested with Claude Desktop (via claude_desktop_config.json), Cursor AI (via Command Panel > Cursor Settings > MCP Server), and VSCode Insider.