Edge appliance · On-prem AI · Layers 1–4

The network engineer that never sleeps.

A small box in your office watches every link, port and service around the clock. When something breaks it finds the real cause and tells you in plain language — without a single packet leaving the building.

Request access See what it does

Runs air-gapped
Vendor neutral
Never inline

stx-edge-01 / Delhi HQ

00.00s

Fleet — 3 sites

Delhi HQ	148 hosts	Investigating
Mumbai	62 hosts	Healthy
Bangalore	39 hosts	ISP fault

Root cause analysis — Delhi HQ

Critical INC-2041 · confidence 96%

Internal DNS server stopped responding

Evidence: Three resolver timeouts in a row. Every layer below transport is healthy.
Impact: Payroll and CRM unreachable for about 40 people.
Do this: Restart the DNS service on SRV-02. No engineer visit, no ISP ticket.

The problem

One fault. Hundreds of alerts. Nobody knows where to look.

The internet slows, VPN drops, printers stop, the ERP goes down. Four teams troubleshoot four different things. Two hours later it was one switch port all along.

One bad port, 100-seat office Illustrative

Measure	Today	With Stellix
Actual fault	One faulty switch port	One faulty switch port
Alerts raised	180+ across four tools	1 correlated incident
Time to root cause	~2 hours, four teams	Seconds, unattended
Who has to read it	A network engineer	Anyone — it arrives as a sentence

What the box does

Detect. Explain. Act — safely.

All of it inside your network. The platform keeps working when the internet doesn't.

01Live

Detect

Watches every router, switch, firewall, AP and server — and catches a link on its way out before it goes.

02Live

Explain

Collapses the storm into one incident, with the readings that prove it, in a sentence anyone can act on.

03In build

Act

Recommends the fix, applies it on approval, and rolls itself back the moment the network gets worse.

Nothing leaves the building. The AI runs on the appliance. Which is why it can go into banks, defence and government sites where cloud tooling simply can't.

Where we play

We own the foundation, not the app

App monitoring is a crowded, solved market. The layers underneath — where outages actually start — are still wide open.

Layer	Name	Scope	What breaks here
07	Application	Not our fight	Someone else's market
06	Presentation	Not our fight	—
05	Session	Not our fight	—
04	Transport	Ours	Is the service actually reachable
03	Network	Ours	Routing, gateways, packet loss
02	Data link	Ours	Loops, storms, VLAN mismatch, MAC flapping
01	Physical	Ours	Cables, fibre, optics, dead ports, heat

Why this one

Built for the network, not the cloud

Vendor neutral

One collector, every box

Cisco, Juniper, Arista, Nokia, Huawei — one event model.

Offline

Runs air-gapped

The AI sits on the appliance. Telemetry stays on site.

Plain language

Answers on WhatsApp

English or Hindi. No dashboard login to know what's wrong.

Proof

ISP evidence reports

Signed and timestamped, so a provider can't wave it away.

Predictive

Warns before it breaks

Flags a link on its way out while there's still time to plan.

Correlated

Network plus the room

Power, heat, humidity and leaks on the same timeline.

Guarded

Automation that reverts

Blast radius checked first. Undone the second it hurts.

Multi-site

Made for MSPs

Every customer office on one screen, at a glance.

The rest

We're not putting the good part on a website

Protocol coverage, the detection models, the hardware and how the evidence chain is signed — that conversation happens in a room, with the box on the table.

Design partners only Under NDA

Protocol coverage
Detection models
Hardware specification
Evidence chain
Build order
Pricing

Ask and we'll walk you through all of it. Request access

Design partners

Get the first box

The first batch is small. We're placing units with offices and MSPs willing to tell us what actually breaks.

A 30-minute briefing, not a sales call
You see everything above, under NDA
Direct line to the people building it

You're on the list

We'll be in touch before the first units ship.