Ask Your NetBox Questions in Plain English
We integrate AI assistants with NetBox using the open-source NetBox MCP server: read-only access to your source of truth, deployed the way an auditor would want it.
What this unlocks
NetBox is where disciplined teams keep the truth about their network: sites, racks, devices, prefixes, IPs, VLANs, circuits. Getting answers out of it still means clicking through the UI or writing API scripts. The Model Context Protocol closes that gap: an MCP server lets Claude and other AI assistants query NetBox directly and reason over the results.
The result feels less like an API and more like a network engineer who has read your entire IPAM: capacity questions, drift questions, "what's at this site" questions, answered in seconds, with the data to back them up.
Built on open source: credit where due
The foundation is the excellent netbox-mcp-server by NetBox Labs (Apache 2.0). We didn't write it. We run it in production, know its edges, and build on it for clients. Its query tools:
netbox_get_objects Retrieve any core NetBox object type with full filter support: sites, devices, prefixes, IPs, VLANs, circuits, racks, and more.
netbox_search_objects Free-text search across object types: "find everything related to den-sw-01" as one question.
netbox_get_object_by_id Pull the complete detail record for a specific object once you know what you're looking at.
netbox_get_changelogs Query the audit trail: who changed what, and when. Drift questions become one prompt.
Questions it answers well
- "Which prefixes in 10.20.0.0/16 still have free space, and what's allocated in each?"
- "List every device at the Denver site that's still marked "staged", and who last modified them."
- "What changed in NetBox in the last 48 hours?"
- "Which VLANs exist at site X but have no assigned prefixes?"
What we add
- Production deployment: stdio for desktop clients or streamable HTTP for shared use, Docker or bare, with read-only API tokens scoped to exactly what the assistant should see.
- Security posture: read-only by default is the right call; we help you keep it that way, and design approval-gated write workflows only where they're genuinely warranted.
- Extensions: custom tools, plugin object types, cross-system integrations (monitoring, ticketing, config data alongside IPAM).
- Beyond NetBox, the same pattern applies to any internal system: we build custom MCP servers that give AI assistants safe access to the tools your team already runs.
Want this running against your source of truth?
A basic deployment is quick; the value is in doing it safely and extending it to fit your environment. Tell us what you run and what you'd want to ask it.
Ready to automate?
Let's discuss your project. I'm available for new engagements.