HYPERLINXWORK · 2026 AutoLinx in production · SecureLinx early access Q3 2026 · FabricLinx ships with both See products →
HYPERLINXWORK · NETOPS + SECOPS PLATFORMS

Intelligent at wire speed.
Evidence at every depth.

HyperLinxWork — platforms for intelligent network operations and security.

HyperLinxWork builds agentic AI for network operations and security — powered by our own reasoning stack that runs on commodity CPU. No GPU lock-in. Three products, one architecture. So engineers never trade speed for evidence.

Our own AI stack · CPU-deployable Engineer-approved by default Self-hosted · air-gap ready 18 years · Bangkok
IN PRODUCTION · 11–16 YEAR RELATIONSHIPS
3 anchor customers · Thailand + Laos
tier-1 telco · national telco · government ministry
Reference calls available under NDA during the briefing process. Request a reference call →
OUR APPROACH

Intelligence needs
a fabric to think on.

Most network platforms make one of two trade-offs. They sample telemetry to keep up — so when something goes wrong, the evidence is already gone. Or they aggregate it into 30-second buckets — so engineers see the symptom thirty seconds late.

We refuse both. HyperLinxWork engineers a single intelligent layer — FabricLinx — that reads your network at full fidelity. Our agents (AutoLinx for operations, SecureLinx for security) reason on top of it. FabricLinx ships with both. It can also stand alone.

That's why our agents read the same network and write into the same audit ledger. NetOps and SecOps stop being two organizations defending two stacks.

One intelligent FabricLinx

full fidelity · wire speed · no sampling

Every packet, every BGP UPDATE, every syslog line, every optical level — read as it happens. FabricLinx scales from a single rack-unit appliance to carrier-grade 100G ingest.

Agents that propose, engineers that decide

agentic AI, security-grade caution

AutoLinx and SecureLinx are agentic by default — but every change carries a diff, evidence, blast radius and rollback. Auto-approval is opt-in, scoped per action class.

Replayable evidence · forever

append-only · signed · vendor-neutral

Every action — from NetOps or SecOps — lands in a single ledger. Export against SOC 2, ISO 27001, or PCI controls. Replay the decision, the evidence, and the context behind it.

DEPLOY ANYWHERE

CPU-first.
No GPU lock-in.

Most agentic AI platforms ship with an implicit assumption: you have GPU capacity, or you can rent it. For the APAC enterprise market — where GPU procurement is hard, datacenter density is constrained, and air-gapped deployments are common — that assumption breaks.

We made the opposite choice. Our reasoning stack — a nano language model paired with graph neural networks — runs on commodity CPU. Pilots on laptop-class hardware. Production on existing servers. Wire-rate workloads handled by FabricLinx (FPGA-based) when you need them. GPU is optional augmentation, never a requirement.

BASELINE

CPU · commodity x86

All three products run on standard server-class CPU. Pilots possible on laptop-class hardware. No special procurement needed.

SCALE UP

FPGA · via FabricLinx

Wire-rate ingest and packet-level forensics handled by FabricLinx on FPGA — from affordable Kria KR260 to carrier-class Alveo. CPU-resident agents stay unchanged above.

OPTIONAL

GPU · for augmentation

Want to augment with frontier LLMs (Claude, GPT, Gemini, vLLM)? Integrate any of them. AutoLinx's baseline reasoning doesn't depend on them — it's an upgrade, not a prerequisite.

PRODUCTS

Three products. One intelligent FabricLinx.

Two agents above. One FabricLinx beneath. Agentic AI is the baseline — across all three.
AutoLinx · NetOps
AutoLinx
Agentic AI for network automation.
Diagnosis, change management and compliance across multi-vendor networks. Every agent action is paired with evidence and gated by an engineer. Reversible by default.
↳ Triage agent · < 60s median diagnose
↳ Change management with rollback
↳ Continuous compliance & audit
↳ 10+ vendors · one intent grammar
For
NOC · network engineering
Deployment
Software · cloud or VPC
Scale
50 – 50,000 devices
FabricLinx
Optional acceleration
In production since 2025 Explore →
SecureLinx · SecOps
SecureLinx
Agentic AI for security observability.
Catch lateral movement, capture packet-level evidence, and detect policy drift across every firewall vendor. Agents propose, analysts decide. No sampling, no aggregation.
↳ Lateral movement · packet-level proof
↳ Policy drift across major vendors
↳ Forensic replay · flow + packet
↳ Sub-100µs detection · FabricLinx-native
For
SOC · Telco · CSP
Deployment
Appliance · on-prem / air-gap
Scale
Carrier-grade · 100G+
FabricLinx
Required · ships included
Early access · Q3 2026 Join early access →
FabricLinx
FabricLinx
Intelligent FabricLinx. See sooner. Respond faster.
The traffic intelligence layer beneath AutoLinx and SecureLinx. Reads every packet at wire speed. Ships from a compact appliance for mid-market to a carrier-grade system for telco.
↳ Full-fidelity ingest · zero sampling
↳ Wire-rate reasoning · microsecond latency
↳ Compact form factor to 100G+ scale
↳ Optical tap or SPAN · inline-free
For
Mid-market to telco
Deployment
Appliance · on-prem
Scale
10G → 100G+
Sold
With AutoLinx / SecureLinx
Ships with both products See specs →
Not sure which?
Most NetOps teams start with AutoLinx as software. Mid-market teams that need wire-rate visibility add FabricLinx. SecOps teams that need packet-level proof — or networks above 25 Gbps — start with SecureLinx. FabricLinx ships beneath either, and many customers run both agents on a single shared ledger.
Talk to engineering →
In production
3 anchor customers · 11–16 years each · Thailand + Laos · regulated industries · AutoLinx evolves from True AutoProvision
Compliance
SOC 2 Type II · ISO 27001 · PCI evidence packs
Data residency
Self-hosted in your VPC or on-prem · no config leaves your network
Our own AI stack
Nano LM + GNN · purpose-built · runs on CPU · no GPU lock-in
COMPANY

HyperLinxWork.

A network platform company, headquartered in Bangkok.

LinxWork was founded in 2008 as a network systems integrator and software house in Bangkok. For nearly a decade we built and operated networks for Thai enterprises and carriers — including AutoProvision, our network automation product, in production at a tier-1 Thai telco since 2017.

In 2022 we began building our own AI stack — a nano language model architecture paired with graph neural networks, purpose-built for the reasoning patterns network operations actually require. We chose to build it ourselves, deployable on CPU, because the alternative — wrapping a foundation model and renting GPU — couldn't reach the APAC enterprise market we'd been serving for 14 years.

In 2026 we rebranded to HyperLinxWork. Same Bangkok HQ. Three products now: AutoLinx (the agent for network operations · evolution of AutoProvision), SecureLinx (the security agent · early access Q3 2026), and FabricLinx (the intelligent traffic layer · for environments that need wire-rate ingest).

Founded
2008 · as LinxWork · rebranded to HyperLinxWork in 2026
AI transformation
2022 · began building proprietary nano-LM + GNN reasoning stack
HQ
Bangkok, Thailand · serving APAC remotely
Team
Network operators · ML researchers · FPGA engineers, working as one team
Customers
3 anchor customers · Thai tier-1 telco · Laos national telco · Laos government ministry · 11–16 years each
Compliance
SOC 2 Type II · ISO 27001 · PCI mapping available
NEXT

Where would you like to start?

Pick the path that fits your team — saves a meeting.