Server Ops MCP

Server Ops MCP

English | 简体中文

[npm version] [License: MIT] [Node.js] [MCP] [GitHub stars]

A general-purpose server-ops MCP Server. Let AI tools like Claude Code / Cursor / Codex manage multiple projects or servers through .mcp.json — log troubleshooting, system resource inspection, code read/write, and Nginx / certificate management.

Supports local and remote (SSH) modes, with built-in two-step confirmation, command whitelisting, path-traversal protection, and automatic secret redaction.

Features

• 🔍 Log troubleshooting: list / read / search project logs, multi-channel logs & system logs
• 📊 Resource monitoring: CPU / memory / disk / load / processes / service status at a glance
• 📁 Code operations: read, write, patch, delete, and search files — all sandboxed to the project root
• 🌐 Nginx & certificates: config test & read, safe reload, certificate install & renewal
• 🔒 Secure by design: two-step confirmation for writes, command whitelist, path-traversal protection, auto redaction

Installation

Requires Node.js 18+. No clone or build needed — npx fetches and runs the published npm package on first use (see the config below).

To build from source instead (e.g. for development):

git clone https://github.com/GT-dinuo/server-ops-mcp.gitcd server-ops-mcpnpm installnpm run build

Build output goes to dist/; the entry point is dist/index.js.

Configuration

In each project where you want to use this tool, create (or append to) .mcp.json and declare an MCP Server per environment. Local mode only needs OPS_PROJECT_ROOT; remote mode adds the SSH variables.

{  "mcpServers": {    "myproject-server": {      "command": "npx",      "args": ["-y", "server-ops-mcp"],      "env": {        "OPS_PROJECT_ROOT": "/www/wwwroot/your-project",        "OPS_SSH_HOST": "<server-ip-or-domain>",        "OPS_SSH_PORT": "22",        "OPS_SSH_USER": "ubuntu",        "OPS_SSH_KEY": "~/.ssh/id_rsa"      }    }  }}

If you built from source, point the command at the local build instead: "command": "node", "args": ["/absolute/path/to/server-ops-mcp/dist/index.js"].

Environment Variables

Usage

Once configured, just give ops instructions in natural language in your AI tool. The AI picks and calls the right tool automatically.

Read-only operations (run directly)

Check the server's CPU, memory, and disk usageSearch for error-level logs from the last hourList the project's log filesRead the Nginx config and check for problemsFind every place in the code that calls sendSms

Write operations (require confirmation)

For writes, deletes, command execution, Nginx reload, certificate install, and similar, the tool first returns a confirmationId. The AI shows you exactly what will run, and it only executes after you confirm:

You: Reload nginxAI: (calls nginx_reload, returns the pending action + confirmationId)    About to run: nginx -s reload. Confirm?You: ConfirmAI: (calls confirm_execute to run it)

file_write, file_patch, file_delete, command_exec, and similar all follow the same confirmation flow.

Typical scenarios

• Production troubleshooting: "Why is the server memory full?" → memory_analysis + log_search working together
• Routine inspection: "Is the disk almost full? Which directory uses the most?" → disk_analysis
• Security audit: "Audit the project config for leaked secrets" → config_audit (output is auto-redacted)

Tool List

System Ops

Project Code

Security

1. Read-only tools run directly.
2. Writes, command execution, Nginx reload, certificate install require two-step confirmation (confirm_execute).
3. Command execution uses a whitelist — dangerous operations like rm -rf, sudo, and piping into a shell are rejected.
4. File operations are sandboxed to OPS_PROJECT_ROOT; path traversal is blocked.
5. Passwords, tokens, and other secrets in .env files and logs are automatically redacted.

Custom Configuration

Download config.example.json as config.json, then point OPS_CONFIG_PATH at it to customize log channels, the command whitelist, read limits, and more.

curl -o config.json https://raw.githubusercontent.com/GT-dinuo/server-ops-mcp/main/config.example.json# (if you cloned the repo: cp config.example.json config.json)# After editing config.json, set in the .mcp.json env:# "OPS_CONFIG_PATH": "/absolute/path/to/config.json"

Main fields of config.example.json:

• logChannels: channel name → log directory path (relative to project root)
• maxReadLines / maxReadBytes: per-read line / byte limits
• commandWhitelist.direct: commands that run directly
• commandWhitelist.confirm: commands that require confirmation

Development

npm run dev    # watch mode, recompiles on changenpm run build  # build to dist/npm run clean  # remove dist/npm start      # run the built Server

Notes

• Remote operations rely on SSH — use a read-only account or a tightly-scoped command whitelist for the AI.
• Private key files should have permission 600 (chmod 600 ~/.ssh/id_rsa).
• Never commit .mcp.json or config.json containing passwords to a repository.

License

MIT © 2026 server-ops-mcp