# shell MCP server

> Enables an AI client to execute allowlisted shell commands with secure argv-based execution, stdin support, and structured audit logging.

- Repository: https://github.com/tumf/mcp-shell-server
- Page: https://hostedmcp.org/servers/tumf/mcp-shell-server
- Transport: —
- Stars: 202
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `ALLOWED_COMMANDS` | no | no |  |
| `ALLOW_COMMANDS` | no | no |  |
| `ALLOW_PATTERNS` | no | no |  |
| `MCP_SHELL_CHILD_ENV_ALLOWLIST` | no | no |  |
| `MCP_SHELL_SAFE_PATH` | no | no |  |

## FAQ

### Is this server a sandbox?

No, it is not a sandbox. Allowing a command delegates that program the server process's existing OS authority, including filesystem and network access.

### How do I restrict which commands can be run?

You can use the ALLOW_COMMANDS environment variable to specify a comma-separated list of allowed executable names, or ALLOW_PATTERNS for regular expression matching.
