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.
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.
What changes for the team
A wider view of the same loop. Each card zooms into the corresponding column above.
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.
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.
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.
Product architecture
Intent in, change plan out.
A virtual twin takes the risk.
Intent ↔ config semantics, plus 30-rule safety net.
Human approves, snapshot taken, then apply — with rollback ready.
Runtime flow
From intent to live device, every change moves through six participants.
Capabilities
By the numbers
- 8 specialized agents
- 30 baseline rules
- 5-level RBAC
- Containerlab + Batfish
- NetBox + Nautobot
- Multi-vendor (5 vendors)