DashyNMSAsk about your LibreNMS network.
“What’s alerting right now?” “How’s the core switch doing?” “What changed on that firewall since yesterday?” DashyNMS MCP lets Claude, or any other AI assistant that speaks MCP, answer questions like these from your own LibreNMS server. It can look at everything and change nothing.
Read-only · Runs on your own machine · Nothing tracked
Three critical alerts, all in the last 40 minutes:
- edge-rtr-02 is down, for 4 minutes
- ups-dc-a is on battery, for 12 minutes
- fw-01 lost a BGP session 38 minutes ago
There are 7 warnings as well, mostly busy ports. The UPS is the one I’d look at first: if it runs flat, edge-rtr-02 won’t be the only thing that’s down.
Up and running in a few minutes
Nothing to install on your LibreNMS server. The MCP server runs next to your AI app, on your own computer or a server of yours, and talks to LibreNMS over its API.
Pull the image
It’s a small container image for Docker, about 45 MB, for Intel and ARM machines alike.
Make a read-only token
Create an API token in LibreNMS for a user who can only read. Then LibreNMS itself would refuse a change, even if something asked for one.
Add it to your app
One command for Claude Code, or a few lines of settings for Claude Desktop. Then just ask.
Step by step instructions are on the support page.
Ask it the way you’d ask a colleague.
Name a device however you like: its hostname, display name, IP address or LibreNMS ID. Every answer comes with the evidence it was based on.
Is it OK?
Up or down, uptime, open alerts, CPU, memory, the fullest disk, and any ports that are down or counting errors, in one answer.
What’s alerting, and why?
Every open alert, or everything that fired overnight. For any alert, what its rule matched, the values that tripped it, and whether it’s a one-off or keeps flapping.
Where is this plugged in?
Give it a MAC or IP address and it follows it to the switch and port it’s on, with that port’s description, VLAN and state. If more than one port could be the answer, it says so rather than guessing.
What happened, and what changed?
A device’s events, alerts and outages for any stretch of time. Connect Unimus and Graylog and you also get the config diff and the log messages from the same window.
How’s the network?
Devices up and down, alerts by severity, the busiest ports, the ones counting errors, and anything running hot, across the whole network or one site.
Show me the graph
Any device’s, port’s or sensor’s LibreNMS graph, as a picture in the chat, from the last hour to the last year.
Safe by design.
An assistant that can look at your network should never be able to change it, or see your secrets. That was the starting point.
It only reads
There’s no tool that changes anything. Use a read-only token as well, and LibreNMS, Unimus and Graylog would refuse a change anyway.
Secrets are taken out first
Passwords, SNMP communities, keys and the like are redacted from config backups before anything reaches the assistant.
Your network’s text stays data
A port description or log line can’t give the assistant orders. Text from your network is passed on as data, never as instructions.
Your tokens stay with you
They’re read from your settings, never logged, and only ever sent to the server they belong to.
More of DashyNMS.
The MCP server is built on the same LibreNMS, Unimus and Graylog code as the apps, so your assistant judges health and alerts the same way they do. Use it on its own, or alongside them.
DashyNMS Desktop
Your LibreNMS, right on your desktop. The full app for Windows, with alert rules, maps and a dashboard you arrange.
See DashyNMS DesktopDashyNMS Mobile
Your LibreNMS, in your pocket. The same alerts, devices and dashboard on iPhone and iPad, in TestFlight beta now, with Android to follow.
See DashyNMS MobileGet DashyNMS MCP
v1.1.0 for Docker, on Intel and ARM. Free to run for your own organisation, with Claude Desktop, Claude Code or any other MCP client. What’s new
docker pull ghcr.io/dashynms/mcp:latest