Network admins are actively exploring AI that targets their day-to-day network operations. These engineers are wiring AI agents directly into their infrastructure, not as a chatbot bolted onto a dashboard, but as a coworker that can query BGP tables, pull switch diagnostics, and reconcile a source of truth, all from a Slack channel.

NetClaw is a rapidly evolving open-source AI network engineering agent built on Anthropic’s Claude, and a prominent example of this fast-growing shift in network management. We’re excited to announce the Lantronix SLC 9000 console server and Percepxion management platform MCP servers and custom AI skill are now officially part of the NetClaw community on GitHub.

As a new edition to NetClaw’s library of hundreds of MCP servers and skills, a network engineer running NetClaw can now ask about console session activity, check out-of-band fleet health, or reliably reach a device, all in the same conversation where they’re already troubleshooting a routing adjacency or auditing a firewall policy. No separate login, no separate tool.

 

A Management Plane of Its Own

Here’s the part that matters most for engineers evaluating AI-driven operations. An AI agent is only as useful as its access to the network. If that access rides on the same production path the agent might be trying to diagnose or repair, the automation has a single point of failure built into it from the start.

Autonomous Out-of-Band access, the approach behind SLC 9000, gives an agent a dedicated, always-available path to every managed device, independent of whatever is happening on the production network. The AI agent doesn’t need to “hope” the production network is available. It has a separate management path specifically designed to remain available when production connectivity fails.

Lantronix Percepxion adds the fleet-wide view on top of that. Instead of checking device by device, Percepxion gives one place to see inventory, health, and access across a distributed OOB deployment, the kind of structured, single source of truth that makes an AI agent more useful and, frankly, more trustworthy.

Where Things Stand Today We want to be direct about something too. NetClaw, and the open-source AI agent movement it’s part of, is genuinely new. It’s a community project growing fast, built by network engineers who are experimenting in labs and small pilots. We recommend evaluating it in a lab or tightly controlled environment before granting it production access. Use scoped credentials, least-privilege permissions, and appropriate change controls. We expect both the tooling and the industry practices around it to mature considerably over the next year or two. Lantronix intends to be part of that maturing, not a bystander watching it happen from a safe distance. Learn More Percepxion and the SLC 9000 already give network teams a resilient, independent way to reach their infrastructure. Now that same access is available inside the tools engineers are using to experiment with AI-driven network operations, on their terms, with the audit trail and access controls a professional network team expects. Learn more about SLC 9000 and Percepxion: lantronix.com/products/slc-9000-series

A Real-World Example Here is one way the addition of Lantronix SLC 9000 & Percepxion increases the resiliency of the NetClaw project: NetClaw detects a BGP adjacency failure. It identifies the affected router. Production management access is unavailable. NetClaw reaches the router through the SLC 9000 console. It retrieves the device state and console output. The engineer gets the diagnosis in Slack. The engineer decides whether to take corrective action.