Product & technology

Calm on the outside. A relentless engine underneath.

You experience a quiet conversation. Behind it, a multi-agent system watches your network every minute, finds the root cause when something drifts, and fixes it from a library of safe, reversible actions — then checks its own work.

The whole product

A five-layer system. The intelligence lives at the top.

The bottom two layers are open and partnered — built on open hardware and open standards (TIP OpenWiFi / OpenLAN), enterprise-grade gear at white-box cost with no vendor lock-in. The top three are where Fortgreene builds. One rule holds the whole thing together: agents may only act on your network through the controller's APIs, which makes every action loggable, reversible and safe.

5

Experience

The conversational interface — your named agent on WhatsApp, chat and voice, with memory and proactive, plain-language updates.

Fortgreene builds
4

Agentic operations

The autopilot itself: sense → diagnose → decide → act → verify → learn. Memory, safety gating and an evaluation harness live here.

Core IP
3

Data & telemetry

Telemetry ingest, time-series and incident stores, an immutable audit log, and strict multi-tenant isolation between sites.

Fortgreene builds
2

Controller

The open OpenWiFi cloud-controller SDK for device management, configuration push and telemetry — the single safe channel through which every action is taken.

Open / partner
1

Hardware

Open white-box access points, switches and a gateway — built on TIP OpenWiFi and OpenLAN, manufactured by an open ODM partner (Edgecore/Accton). Enterprise-grade at white-box cost, multi-vendor, with embedded cellular and IoT radios (BLE, Zigbee, Thread) for the physical-AI era. No lock-in.

Open · white-box

The control loop

Six steps, repeated forever.

A supervisor agent orchestrates specialist agents over a shared, stateful picture of your network. Diagnosis may use LLM reasoning — but the model only ever chooses which bounded action to run. It never writes one.

01 · Observe

Sensing raises a candidate

Continuous telemetry surfaces the first sign that something has drifted from healthy — long before it becomes an outage you'd feel.

02 · Orient

Diagnosis finds the cause

The signals are correlated into a likely root cause with a confidence score — the difference between "something's off" and "this AP's channel is congested."

03 · Decide

It selects a bounded action

It picks one action from a coded library and checks the gating rules to decide whether it's safe to run automatically.

04 · Act

It executes via the controller

The action runs through the controller's APIs — or pauses for your confirmation if it's high-impact. Never a raw, invented command.

05 · Verify

It confirms the fix held

Fresh telemetry checks the problem is actually gone. If it isn't, the loop continues or the incident escalates to a human operator.

06 · Learn

It feeds the flywheel

The full trace joins the evaluation set and sharpens future decisions. Every network Fortgreene runs makes every network it runs a little smarter.

The roster

Not one model. A team of specialists.

A multi-agent, multi-LLM engine — built on the founders' proven multi-agent heritage — where each agent owns one job and a supervisor keeps them coordinated and safe.

Orchestrator
Supervisor

Owns the loop, routes work to the right specialist, and enforces the safety gating on every action.

Sensing & telemetry
Monitoring

Watches the live stream from every device and raises a candidate incident the instant something looks wrong.

Diagnosis
Root-cause analysis

Correlates telemetry into a root cause and a confidence level, so the fix targets the real problem.

Remediation
Healing

Selects a bounded action, executes it through the controller, and verifies the result before moving on.

Security
Zero-trust enforcement

Discovers and classifies every device, segments traffic, and quarantines anything it doesn't recognise.

Owner conversation
Natural language

Translates your intent into network state and back into plain language — and reaches out proactively when it matters.

The boundary that matters: agents may act on the network only through the controller's APIs. That single rule is what makes every action loggable, reversible and safe — the model decides which action, never invents a new one.

Two surfaces, by design

The owner gets simplicity. Operators keep full control.

Dashboards aren't eliminated from the system — they're simply not for you. Setup itself is a roughly two-minute plain-language conversation that auto-generates the entire network; the configuration lives where it belongs.

Surface 1 · the owner

The agent on your phone

The only thing you ever see. A named agent over WhatsApp, SMS, voice, email or chat — no dashboard, no console, no config. You describe your business in a short conversation and the network builds itself: SSIDs, secure zones, guest access and zero-trust roles, all generated for you.

Surface 2 · the operator co-pilot

The glass box behind the curtain

The internal console used by Fortgreene operations, partners and MSPs — never the owner. It holds the auto-generated config, the pending Tier-B approvals, the bounded action library and an immutable, append-only audit log. Full control and accountability, kept out of your way.

See the engine run on your own floor.

Early access is open to founding sites. Start with one location and watch the autopilot handle a real day.