What's next
You've worked through the architecture. Here's where to go from here.
Build one for real — the Weekend Sprint
Reading the architecture is one thing. Shipping an agent against your own network is another. The Ship Your First AI Agent — Weekend Sprint closes that gap. You bring a real NOC/SOC task you're sick of doing by hand. You leave — one weekend later — with a working agent that handles it, on your stack, with a deployment path your security team will sign off on.
Small group. 15 seats. Founding price €149.
Advanced topics not covered
- Multi-vendor parity — Cisco IOS-XE and Arista EOS have the strongest representation in Claude's training data. Junos, PAN-OS, Fortinet, and Nokia SR Linux are supported but with less consistent accuracy. If your network runs primarily on one of those platforms, your CLAUDE.md and skills become even more important — they supplement Claude's weaker areas with explicit operational knowledge.
- Verify-before-trust — Claude knows a tremendous amount about network protocols and vendor CLI syntax. It also occasionally generates commands with subtle errors: a wrong keyword for a specific software train, a misremembered default timer value, a feature that exists in one EOS version but not another. Validate every command before production use. The tool accelerates the production of correct answers. It does not guarantee them.
- Custom MCP servers — If no MCP server exists for your specific monitoring platform (LibreNMS, Observium, etc.), someone needs to build one. The MCP specification covers this; it's a small Python or Node project that wraps your platform's API.
- The Agent SDK — For teams building products on top of Claude Code (custom NOC platforms, internal tools that orchestrate Claude programmatically), the Agent SDK is the next layer down. Not needed for daily operational work.
Resources
Community
- Anthropic community Discord — where the Claude Code power users hang out
- NetBox community — if you're rolling out the IPAM-as-source-of-truth pattern, the NetBox Slack and GitHub discussions are the best signal
For your NOC
If you've worked through all twelve sections, you have the architecture. The remaining work is making it yours: your vendor stack in CLAUDE.md, your operational procedures as skills, your safety rules as hooks, your team's tooling as a plugin. If you'd rather not derive that from first principles, book a call — we'll skip the trial-and-error and design it for your specific NOC.