Skip to content
Cloud
Sign up

Aspian Cloud keeps your systems in one jurisdiction, keeps secret values away from people and agents, and records every change in an audit trail you can verify without asking us.

Each cell is a self-contained cluster in one jurisdiction.

Your apps, databases, logs, backups and audit records stay inside the cell you choose. The cell holds everything it needs to keep running, so if our central control plane is unreachable, your apps keep serving. Only new deploys pause.

Where cells run and where their backups go
CellBackups go to
LagosFor Nigerian data.A second Lagos facility
AbujaFor Nigerian data, close to federal government.A second facility in Nigeria
LondonFor UK data.A second UK provider
European UnionFor workloads with no residency requirement.Separate storage in the EU
Your data centreThe same platform on your hardware, run by Aspian.Storage you choose, in your jurisdiction

Control plane

Unreachable

Deploys, plans and dashboards

Lagos cell

Self-contained cluster, one jurisdiction

Serving
  • AppsServing traffic
  • Databases and backupsRunning
  • Logs and audit recordsWritten locally
  • New deploysPaused until the link returns

Every store is listed, including the ones outside the country.

Data is stored where you agreed. Some services process it elsewhere, and we name each one rather than leave it out.

WhatsApp and email providers process messages outside Nigeria. For Nigerian organisations, those messages carry only an alert code and a secure link, never the details. If you connect an AI coding agent, what it reads goes to its model provider.

Where guard-roster’s data is stored and processed

Kings Guards, production

Stored in Lagos
App and databasesLagos cell, primary facilityObservedToday 10:00
BackupsSecond Lagos facilityObservedToday 06:00
Logs and build artefactsLagos cellObservedToday 10:00
Audit recordsLagos cell, anchored to write-once storageObservedToday 10:00
SMS alertsNigerian SMS providerConfiguredAwaiting verification
WhatsApp and email alertsProviders outside NigeriaDisclosedAlert code and link only
AI coding agent via MCPModel provider, outside NigeriaDisclosedLogs it reads leave the cell
Observed: measured by the platform. Configured: set, not yet measured. Disclosed: processed outside Nigeria, and named in your data processing agreement.

Secrets go in. Their values don’t come back out.

Secrets are stored in OpenBao with a write-only policy. No dashboard, API, CLI command, MCP tool, log or export shows a stored value.

AI agents get a handle they can pass to a deploy, not the value. Only the running app receives it, scoped to its own environment.

DATABASE_URL

Write-only
Dashboard, CLI and API
Name, version 3, updated by D. Eze on 2 Nov. Replace or delete only.
AI agent through MCP
A handle it can pass to a deploy: secret://guard-roster/production/DATABASE_URL
Logs and exports
The key name, never the value
The running app
The value, at runtime, scoped to this app and environment

Other customers can’t reach your apps or your rows.

Isolation doesn’t rest on one control. Each layer assumes the one above it might fail.

Sandboxed runtimes
Every app runs in a gVisor or Kata sandbox, so code that escapes its container still meets another boundary before the host.
Dedicated nodes
Regulated organisations get machines of their own inside the cluster. Their workloads don’t share a node with another tenant.
Row-level security
In the platform’s database, every tenant-owned table carries a policy that limits each request to one tenant’s rows.
Signed images only
The cluster admits only images built and signed by our pipeline. Unsigned images, or images signed by another key, are refused.

An audit trail that shows if anyone has changed it.

Each tenant has its own chain. Every entry carries the hash of the entry before it and an Ed25519 signature from a key held outside the app.

The database refuses edits and deletes. A database administrator could still try, so every hour the chain head is copied to write-once storage outside the database, and the verifier checks against it.

  1. Entry 4,817

    Signed

    Approval by T. Adeyemi

    Person, passkey

    previous
    9f2c…a41e
    this entry
    1b7d…03c9
  2. Entry 4,818

    Signed

    Approval by K. Okafor

    Person, passkey

    previous
    1b7d…03c9
    this entry
    e45a…7f10
  3. Entry 4,819

    Signed

    Migration applied

    System, plan 7Q2F

    previous
    e45a…7f10
    this entry
    6c08…d2b4

Chain head 6c08…d2b4 anchored to write-once storage

11:00, next anchor 12:00

Changing any entry breaks every hash after it, and the anchor outside the database shows where.

The riskiest actions need more than one person.

That includes us. Aspian staff go through the same plans, approvals and audit trail as your own team.

Passkeys bound to the plan
An approval is a passkey signature over the exact plan the approver reviewed. If the plan changes, the approval no longer counts.
Two-person control
Regulated work needs two named approvers. Neither can be the person who asked, and one person can’t approve twice.
Operators work through runbooks
Our staff and your trained operators act through named runbooks and narrow grants, scoped to a tenant, app and environment.
Break-glass is the exception
Direct access to a cluster needs two approvals, expires on a timer and is recorded in full.

A way out that we test.

We run export drills: the bundle is booted on a clean machine and a clean cluster with no Aspian components. The date and result of the last drill appear on your Trust Page.

The bundle includes

  • Your apps as Terraform, Helm charts and Docker Compose
  • Database dumps you can restore with standard tools
  • Your runbooks, operator training and authorisation records
  • Your incident and change history

Secrets come back as a list of keys to re-enter. Credentials, DNS and third-party services still need transition work, and the bundle lists it.

Bring us the system you can't afford to get wrong.

We are opening Aspian Cloud to organisations in stages. Create an account and we will set up your workspace, or talk to us about the systems you run today.