netmiko-mcp

# Netmiko MCP Server > **WARNING:** You can make serious, incredibly detrimental mistakes by using this tool. This tool could cause massive outages in your environment. You, and you alone, are solely responsible for using this tool. Don't say I didn't warn you. > I have tried to make reasonable defaults and to limit what the Netmiko-MCP server allows (by default). It is highly advisable to start with ONLY show commands executed against ONLY test or lab devices. You should also strongly consider additional security mechanisms completely outside of the LLM and Netmiko- MCP (for example, a tightly-controlled AAA solution). LLMs and LLM-agents inherently have a lot of variance and are difficult to predict and control. <br /> ## Welcome to an LLM-assisted Method of Interacting with your Network netmiko-mcp is a Model Context Protocol server that gives your AI client (Claude, Claude Code, Cursor, VS Code Copilot, and others) direct, controlled access to your network devices via SSH (and Telnet if absolutely necessary). Ask a question in plain English; the agent figures out which devices to query, runs the permitted commands, and returns results as a conversation, a formatted table, or structured JSON. Whatever fits your workflow. This repository is a starting point for network engineers curious about what AI agents can do when given access to real devices. The setup process itself is AI-assisted: the `skills/` directory contains reference files that load directly into your AI client's context, so you can ask your AI-client to help you configure the server, set up inventory, or troubleshoot a connection. The skills should have the relevant details. <br /> ## Security Recommendations The controls in Netmiko-MCP are a best-effort layer and should not be your only line of defense. You should strongly consider using AAA (e.g. TACACS+) with a dedicated read-only service account. AAA should independently authorize and audit what commands the account can execute on you…