NAPT MCP Server — Control Your Network From Claude
Your AI assistant, wired directly into your network automation platform.
The Model Context Protocol (MCP) lets an AI assistant call real tools. The NAPT MCP server exposes the platform — templates, script generation, intent parsing, dry-runs, pipelines, and findings — so you can drive your network from a chat window.
How it fits together
Your assistant talks to the NAPT MCP server, which talks to NAPT, which talks to your devices. Every call is scoped by your MCP key, so the assistant only ever sees what you can see.
Step 1 — Get a key
Open the MCP Keys panel and mint a key. Keep it read-only unless you actually plan to run scripts — execution permission is opt-in for a reason.
Step 2 — Add NAPT to your client
For Claude Desktop or Cursor, add the server to the MCP config:
{
"mcpServers": {
"napt": {
"command": "npx",
"args": ["-y", "@napt/mcp"],
"env": { "NAPT_MCP_KEY": "napt_mcp_..." }
}
}
}Step 3 — Test it
Restart your client and ask: "List NAPT templates for BGP." The assistant calls napt_search_templates and returns vetted scripts. From there you can generate, dry-run, and review findings without leaving the conversation.
execute only on the key you use for change windows.Was this helpful?
Generate scripts for the task covered in this post
Describe what you need and NAPT writes production-ready automation in Netmiko, NAPALM, or Nornir.
Open Script Generator →Related posts
Set Up NAPT in 30 Minutes: From Empty Terminal to Your First Automated Check
A guided install: create a Python virtual environment, install NAPT, wire up SSH credentials, and run your first read-only check with --dry-run safety on.
Read →Build a Network Automation Lab: GNS3, EVE-NG, and DevNet Sandbox
Three ways to get real devices to automate against — local emulation, browser-based topologies, and Cisco's free cloud sandboxes — wired up to NAPT.
Read →Building Network Workflows Visually: Chain Templates Into Repeatable Automation
Drag templates onto a canvas, connect them with edges, add branches and parallel groups, then save, schedule, and run the whole flow.
Read →