On-Premise · Self-Hosted · Air-Gapped
Alvor is a security architecture management and compliance platform that deploys as containers on your infrastructure, works fully air-gapped, and runs its AI on models you host. Single-tenant in every model, because it was built that way.
Your boundary
Eight modules, one record, full audit trail
Deployment Models
Every model is the same containerised platform. Choose where it lives; nothing else changes.
A single-tenant instance in the region you choose, with its own database. We run it: managed backups, automatic updates, no co-tenants, ever.
The same containerised platform, deployed on your own servers, in your data centre or your own cloud accounts. Nothing changes but where it runs.
A fully disconnected deployment for environments where nothing crosses the boundary. The AI Assistant points at a model you host inside it.
Dedicated cloud
On-premise
Air-gapped
| Model | Who operates it | Where data lives | Network path | AI model |
|---|---|---|---|---|
| Dedicated cloud | Alvor | Your chosen region, your own database | Internet, to your users | Your key, your provider and region |
| On-premise | Your team | Your data centre or your cloud accounts | Your network | Any provider, or a model you host |
| Air-gapped | Your team | Inside the boundary | No outbound connections | A model you host inside the boundary |
Where it runs
Each circle is a module and each arrow is a link the product makes today. Everything inside the dashed line runs in your instance.
Alvor never bundles a model. Connect Anthropic, OpenAI, Google, Azure OpenAI, or Amazon Bedrock, or point the assistant at any OpenAI-compatible endpoint: a vLLM server on your GPUs, an Ollama instance in your data centre. The agentic studios work exactly the same, the AI proposes and a person approves, and in an air-gapped deployment no prompt or record ever leaves your network.
Model provider
Why Alvor
One shared control plane serving every customer. There is nothing to hand you: your records live in their tenant, behind their boundary, and “on-premise” would mean rebuilding the product. That is why the mainstream compliance platforms do not offer it.
Built single-tenant from the start. Your instance is already a self-contained set of containers with its own database, so on-premise is a deployment decision, not a rewrite. The cloud product, the on-premise product, and the air-gapped product are the same product.
Who Runs Alvor On-Prem
Classified networks and clearance regimes rule out shared infrastructure by policy, not preference. An air-gapped Alvor deployment lives where the work lives.
Sovereignty obligations and procurement rules increasingly require the platform, the data, and now the AI to run on approved infrastructure inside the jurisdiction.
Operators of essential services need their security system of record available when the internet is not, and inside the same protected environment as the systems it governs.
Data residency, outsourcing registers, and vendor concentration limits get simpler when the answer to "where does it run?" is "on our own servers."
Deployment, answered
Best for
Organisations that need the whole security program, including its AI, to run on infrastructure they control.
All three are the same containerised platform with the same eight modules and the same AI Assistant. The difference is where it runs and who operates it. A dedicated cloud instance runs in the region you choose, with its own database, operated by Alvor. An on-premise deployment runs on your own servers or in your own cloud accounts, operated by your team. An air-gapped deployment is the on-premise platform with no path to the internet. Every customer is single-tenant in every model: one instance, one database, no co-tenants.
Nothing has to. The platform, its database and its audit trail run inside your boundary. The AI Assistant sends prompts only to the model endpoint you configure, so pointing it at a model hosted inside the same network keeps prompts, diagrams and evidence inside as well. In an air-gapped deployment there are no outbound connections at all.
Any model served through an OpenAI-compatible endpoint. That includes a vLLM server on your own GPUs and an Ollama instance on a workstation or server. Cloud providers work the same way, with your own key: Anthropic, OpenAI, Google, Azure OpenAI and Amazon Bedrock. Whichever model you choose, the AI proposes and a person approves, and every action is audit-logged. Recreating a diagram from a photo needs a vision-capable model.
Defence and national security programs whose networks are classified. Government agencies with sovereignty obligations that cover the platform, the data and the AI. Critical infrastructure operators who need the security system of record available when the internet is not. Regulated banks, insurers and health providers whose outsourcing registers get shorter when the platform runs on their own servers.
Questions
Most cannot: mainstream compliance and GRC platforms are multi-tenant SaaS, and there is no version of them to hand over. Alvor is single-tenant by design, so the same containerised platform that runs as your dedicated cloud instance deploys onto your own servers, with all eight modules, the AI Assistant, and the audit trail intact.
The platform ships as containers that run on your infrastructure, in your data centre or your own cloud accounts. It is the same product as the cloud version, so the feature set, the record model, and the interface are identical; you keep your own keys, retention policies, and backups inside your boundary.
Yes. Alvor supports fully disconnected operation for defense, government, and critical-infrastructure environments. No outbound connections are required, and the AI Assistant keeps working by pointing at a model hosted inside the boundary.
Yes. Alvor is bring-your-own-model, and any OpenAI-compatible endpoint qualifies as a provider, including a vLLM server or an Ollama instance running on your own hardware. The agentic studios behave exactly as they do with a cloud provider: the AI proposes, a person approves, and everything is audit-logged.
No. Every cloud customer gets a dedicated single-tenant instance with its own database, in the region they choose. There are no co-tenants in any deployment model.
In practice they describe the same decision: the software runs on infrastructure you control instead of the vendor's. On-premise usually means your physical data centre, while self-hosted includes your own cloud accounts. Alvor supports both through the same containerised deployment, and adds a third option, fully air-gapped, for environments where nothing crosses the boundary.
Get started
Whether you lead security, run IT, manage compliance, or sit in the C-suite - we'll show you your view.