Linux Info Mcp

What is Linux Info Mcp?

Linux Info Mcp provides over 60 read-only diagnostic tools that let Claude inspect your Linux fleet files, systemd, performance, networking, processes, disks, kernel, packages, and Docker via SSH. Each tool is a hardened wrapper around a standard command, runs over your own SSH config, and is designed for system administrators and developers who want safe, AI-powered observability of production systems.

How to use Linux Info Mcp?

Set up with Python via uv, configure with a single mcpServers entry in your MCP client, and tune behavior using environment variables. Works with Claude Desktop and Claude Code. Run any tool on one host or pass a list for parallel execution (capped at 25 hosts per call).

Key features of Linux Info Mcp

• Read-only by construction: no code path mutates state.
• Injection-hardened: shell quoting, flag whitelists, host allowlist.
• Uses your existing SSH config and agent; no credentials shared.
• Bounded output, per-call timeouts, capped multi-host fan-out (25 max).
• Auditable: structured stdout/stderr/exit_code + optional JSONL logs.
• 62+ hardened diagnostic tools covering system aspects and Docker.

Use cases of Linux Info Mcp

• Inspect systemd services across multiple hosts without manual SSH.
• Check disk usage and kernel parameters on a fleet of Linux servers.
• Troubleshoot network performance using bounded, auditable SSH commands.
• Monitor Docker containers in production with no risk of mutation.
• Audit package versions across your Linux infrastructure via Claude.

FAQ from Linux Info Mcp

How is this different from other server monitoring tools?

It provides read-only access with no mutation path, injection-hardened commands, and uses your own SSH config – no agent or credentials stored. It is designed for AI-driven inspection via the Model Context Protocol.

What are the runtime requirements?

Python via uv; uses your existing SSH setup (agent, jump hosts, wrapper). No server-side installation is needed on target hosts.

Where does the data live?

All commands run over SSH on the target hosts; outputs are returned to Claude. The server itself never stores credentials or data. Optional JSONL logging can be enabled for auditing.

What are the known limits?

Output is byte-capped, per-call timeouts apply, and multi-host fan-out is capped at 25 hosts to prevent overload. No mutation path exists – no service restarts, file edits, or package installs.

What transports and authentication does it support?

Uses your existing SSH agent or any wrapper (e.g., tsh ssh, jump hosts). The server never sees a credential; authentication is handled entirely by your SSH configuration.