← All projects
Network

NetSage

AI network engineer

Overview

NetSage is an AI-driven platform that takes over the network engineer workload: design, config generation, troubleshooting, deployment, and audit. Eight specialized agents collaborate under a three-gate pipeline — every change passes Containerlab simulation, Batfish validation, and human approval with snapshot rollback before it touches a real device.

It speaks multi-vendor (Huawei VRP, Cisco IOS-XE, H3C, Juniper, Arista) and reads source-of-truth from NetBox or Nautobot. The 30-rule baseline (CIS + vendor hardening) catches the dangerous stuff automatically.

Product impact

Pain → Solution → Value

Network operations depend on deep vendor-specific knowledge that few engineers actually hold — and a single misconfig can take down a region. NetSage replaces tribal CCIE work with a coordinated team of 10 specialized AI agents, a three-gate engine, and multi-vendor support out of the box.

Pain
Vendor-locked config
Huawei / Cisco / H3C each speak their own.
Solution
10 agents + multi-vendor
Huawei VRP, Cisco IOS-XE, H3C, Juniper, Arista.
5
Vendors, 1 workflow
Pain
Misconfigs cause outages
One typo can take down a region.
Solution
Three-gate engine
Simulate → validate → approve + rollback.
0
Unintended outages
Pain
Scarce CCIE engineers
Few hold the tribal knowledge.
Solution
Planner + containerlab sim
Topology-driven deploys.
↓" data-i18n-zh="">
Days → hours
Pain
Tribal troubleshooting slow
Outages sit until a CCIE is online.
Solution
10 named agents on-call
planner, validator, troubleshooter, …
24/7
Always-on network engineer

What changes for the team

A wider view of the same loop. Each card zooms into the corresponding column above.

Pain

Network ops depend on tribal CCIE knowledge

  • Deep vendor-specific knowledge — Huawei VRP, Cisco IOS-XE, H3C, Juniper, Arista — held by few engineers.
  • A single typo in an ACL or BGP config can take down a region.
  • Troubleshooting is slow, tribal, and dependent on who's online.
5
Vendors each with its own dialect
Solution

10 agents + a three-gate engine

  • Specialized agents — planner, config_engineer, validator, troubleshooter, observer, security_auditor, … — each own a stage.
  • Containerlab simulation + Batfish validation + human approval with snapshot rollback before any device is touched.
  • Multi-vendor out of the box: Huawei, Cisco, H3C, Juniper, Arista.
10
Specialized agents
Value

Deploys in hours, outages become rare

  • Changes that took days ship in hours — the planner + simulator pipeline replaces tribal pre-checks.
  • Misconfig-driven incidents drop sharply: every change is simulated, validated, approved.
  • One platform runs the whole multi-vendor network — no tribal knowledge required.
3
Gates before any device is touched

Product architecture

01
Plan

Intent in, change plan out.

02
Simulate

A virtual twin takes the risk.

03
Validate

Intent ↔ config semantics, plus 30-rule safety net.

04
Approve + apply

Human approves, snapshot taken, then apply — with rollback ready.

Runtime flow

From intent to live device, every change moves through six participants.

Technical architecture

AI network engineer. Click any node or relationship for details.

Capabilities

Three-gate engine
Containerlab simulation → Batfish validation → human approval + snapshot rollback.
Multi-vendor
Huawei VRP / Cisco IOS-XE / H3C / Juniper / Arista.
Source of truth
NetBox + Nautobot dual adapters.
Compliance audit
30 baseline rules (CIS + vendor hardening) + ACL analysis.

By the numbers

  • 8 specialized agents
  • 30 baseline rules
  • 5-level RBAC
  • Containerlab + Batfish
  • NetBox + Nautobot
  • Multi-vendor (5 vendors)

Try it, fork it, or hire me to extend it.

Source on GitHub · {p['license']} · questions to 286043314+echocc00@users.noreply.github.com