MCP.md

MCP Integration

SecureShell runs as an MCP server for Claude Desktop, Cline, and other MCP clients.

Quick Start

1. Configure Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "secureshell": {
      "command": "python",
      "args": ["-m", "secureshell.integrations.mcp_server"],
      "env": {
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

See mcp_config.json for examples with all providers.

2. Restart Claude Desktop

The execute_shell_command tool will be available.

Programmatic Usage

See cookbook/mcp_example.py for using the MCP server with your own OpenAI agent.

Supported Providers

Auto-detects from environment:

• OPENAI_API_KEY → OpenAI
• ANTHROPIC_API_KEY → Anthropic
• GEMINI_API_KEY → Gemini
• GROQ_API_KEY → Groq
• DEEPSEEK_API_KEY → DeepSeek

Security

All commands go through:

1. Risk classification
2. Sandbox validation
3. Gatekeeper LLM evaluation
4. Audit logging