@tumf
Basic information
Secure MCP server for whitelisted shell command execution with stdin, argv pipelines, timeouts, and structured audit logging.
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"mcp-shell-server": {
"command": "npx",
"args": [
"-y",
"@smithery/cli",
"install",
"mcp-shell-server",
"--client",
"claude"
]
}
}
}Tool list not available yet
This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.
Fetch the live tool list by running this server in a temporary sandbox using the button above.
A secure shell command execution server implementing the Model Context Protocol (MCP). It allows remote execution of a pre‑whitelisted set of shell commands with optional stdin input, designed for controlled, auditable command‑line access from MCP clients.
Install with pip install mcp-shell-server or automatically via Smithery (npx -y @smithery/cli install mcp-shell-server --client claude). Set the ALLOW_COMMANDS (or ALLOWED_COMMANDS) environment variable with comma‑separated command names, then start the server with ALLOW_COMMANDS="ls,cat,echo" uvx mcp-shell-server. Configure the MCP client (e.g., Claude Desktop) with a JSON entry pointing to this command and any required environment variables.
<, >, >> to the working directoryls, grep, or catOnly command names listed in the `ALLOW_COMMANDS` (or `ALLOWED_COMMANDS`) environment variable, or matching a full‑match regex from `ALLOW_PATTERNS`, are allowed. The server also rejects known exec‑capable vectors like `find -exec` or shell interpreters even if the command name is allowlisted.
Children receive a minimal environment (only `PATH` and essential Windows keys by default). Additional variable names can be inherited from the parent or per‑call overrides using `MCP_SHELL_CHILD_ENV_ALLOWLIST`. Secret‑like names (e.g., containing `TOKEN`, `PASSWORD`) are redacted in logs.
The default timeout is 30 seconds, the maximum accepted timeout is 300 seconds, and the output limit per stdout/stderr stream is 1 MiB. Client‑supplied timeouts are clamped to the server maximum; processes exceeding these limits are terminated and reported.
No. It is not an OS‑level sandbox. While it enforces a command allowlist, argument hardening, and environment isolation, allowed binaries can still read accessible files or consume CPU. For untrusted workloads, run the server inside an external container, VM, or OS policy boundary.
The server is designed for stdio transport, as shown by the MCP client configuration examples that run it as a subprocess (e.g., via `uvx`). It communicates with the MCP client over standard input/output.
Know when your AI agents, cron jobs and CI pipelines fail, stall or go quiet. Your agent sets up monitors from a plain description, tells you what is broken and why, reads run history and traces, and writes notes on inci
Agent-Native Terminal
MCP server that lets Claude Code, Cursor or Codex ask you for API keys on a local page and write them to .env. The agent only ever sees key names.
SSH MCP server for AI agents: remote commands, file transfer, log search and server audits through OpenSSH.
Catch intent conflicts before code conflicts. The open-source coordination protocol for coding agents, built above Git.
Comments