Platform & Deployment

On-prem deployment

Live

Air-gapped install via Helm chart. No outbound telemetry when AIR_GAPPED=true.

Plan availability. On-premise is not part of any published plan and cannot be self-served. None of Starter, Pro, Business or Enterprise switches it on; it is enabled for your organisation under a written agreement covering licensing, support and updates. Talk to us before you plan around it.

Try LexVio free →Open scanner

What you get

✓Helm chart with API, worker, virus-scanner, ingress, autoscaling and disruption-budget templates
✓Air-gapped mode on by default, enforced by refusal in the application code at each outbound vendor
✓RS256-signed licence heartbeats, signature verified before any lookup
✓A backup schedule with defined daily, weekly and monthly retention in the chart

+1 more capabilities below

Overview

What it is.

For organisations that cannot use SaaS — government, defence, regulated finance — LexVio ships as a Helm chart that runs inside your Kubernetes, with the API, the worker pool and the virus scanner as deployments. Licensing is by RS256-signed heartbeat, and the signature is verified before any database lookup.

With air-gapped mode on, and it is on by default in the chart, the outbound paths are closed by hard gates in the application code, not by configuration convention: Anthropic, OpenAI, Google, the speech vendor, the document parser, the jurisdiction verification adapters and error monitoring each refuse to initialise. Worth being precise about the consequence — those gates mean no model inference at all, not model inference served locally.

How it works

Three steps.
End to end.

1. Sign enterprise contract

On-prem requires a written agreement covering licensing, support, and updates.

2. Helm install

We provide the Helm chart and ship updates via signed releases. You install in your own Kubernetes.

3. Heartbeats for licensing

Periodic RS256-signed heartbeats validate the licence. No customer data leaves.

Capabilities

What you get.

✓Helm chart with API, worker, virus-scanner, ingress, autoscaling and disruption-budget templates
✓Air-gapped mode on by default, enforced by refusal in the application code at each outbound vendor
✓RS256-signed licence heartbeats, signature verified before any lookup
✓A backup schedule with defined daily, weekly and monthly retention in the chart
✓Deployment, onboarding and signed updates handled by our engineering team under contract

FAQ

Quick answers.

Can we point it at our own self-hosted LLM?

Not today. The Anthropic client is constructed without a configurable base URL and there is no OpenAI-compatible endpoint setting, so there is nothing to route to a local model server. The chart carries a disabled block for a local inference server that nothing in the application talks to. Do not plan on this until we tell you it is done.

What does air-gapped actually mean here?

That every outbound integration refuses to start — model vendors, speech, document parsing, jurisdiction verification and error monitoring. It is enforced in code rather than by network policy, and the chart does not currently ship an egress network policy, so belt-and-braces network controls are yours to add.

Does the product still work air-gapped?

The parts that do not need a model do. Anything requiring inference does not, because there is no local model path yet. That is the trade air-gapping makes today.