Bring Your Own Cloud: KiviDB in Your AWS, Google Cloud or Azure Account

Abbas Ali ·

KiviDB databases can now run in a cloud account you own. Connect an AWS account, a Google Cloud project or an Azure subscription, and KiviDB creates your databases there: the machines, the disks and the backups live in your account and appear on your cloud bill. KiviDB keeps operating them, with failover, recovery, backups and upgrades, from the same console you already use.

Bring your own cloud (BYOC) is available today, on all three clouds, for every organization.

Settings → Cloud accounts with three verified connections: an AWS account, a Google Cloud project and an Azure subscription.

Why run it in your account

  • Your data never leaves your account. Disks, snapshots and exports are resources you own, in regions you choose.
  • Your network, your rules. Databases sit in your own subnets, so your applications reach them over private addresses inside your VPC, VPC network or virtual network.
  • Your cloud bill. The machines and storage are billed by your provider, so they count toward your commitments and credits.
  • Still managed. You don't run the database. KiviDB still handles failover, recovery, persistence, backups, upgrades, metrics and logs.

What lives where

Everything that holds or serves your data is in your account:

  • the machines and their disks
  • each database's public address and network rules
  • the groups that run read replicas on Pro: Auto Scaling groups on AWS, managed instance groups on Google Cloud, virtual machine scale sets on Azure
  • disk snapshots, and KDB exports in a bucket you choose

KiviDB runs the console and API, the DNS names your applications connect to, the TLS certificates, and the metrics and logs you see in the console. Every resource KiviDB creates in your account is tagged (or, on Google Cloud, labelled) managed-by=kividb, so you can always find it.

Access that's narrow and revocable

KiviDB never gets broad access to your account, and on every cloud there's a clear way to take it away.

  • AWS: you create an IAM role from a CloudFormation template KiviDB provides. It works only in the regions you choose, can change or delete only resources KiviDB created, can touch only the kividb/ prefix of your backup bucket, and cannot create users, roles or policies. Delete the stack to revoke it.
  • Google Cloud: a setup script or Terraform module creates a service account, kividb-mgmt, with custom roles. KiviDB acts as it by impersonation, so no key is ever created. Its permissions are limited to resources named kividb- wherever Google Cloud allows it, and to the kividb/ prefix of your bucket.
  • Azure: KiviDB is a Microsoft Entra app, added to your tenant with one consent. A setup command or Terraform module gives it management of one resource group you choose and nowhere else, read access to the subscription, join on the subnets you name, and the kividb/ prefix of one storage container. No credential of yours is created.

Permissions and security lists exactly what each one can and cannot do.

Connecting takes a few minutes

Open Settings → Cloud accounts and click Connect a cloud account. A guided flow asks for the account details, the regions and subnets you want databases in, and a backup bucket. It then gives you the command or template that grants access. Your progress is saved after each step.

The last step is Verify. KiviDB checks everything a database will need before you create one: that it can sign in, that it has every permission it needs, that your policies allow the regions and resources it will create, that each region's network can reach the internet, that you have vCPU quota, and that it can write, list and delete under kividb/ in your bucket. Anything wrong comes back as a specific instruction, not a generic error.

The Verify step for an Azure subscription, with every check passing: sign-in, subscription, resource providers, the resource group, permissions, Azure Policy, the network, vCPU quota, backup storage and browser uploads.

Creating a database

Nothing about creating a database changes, except one choice at the end. Pick the cloud, region, plan and size in the configurator as usual. Then, on the review page under Where it runs, choose your connected account instead of KiviDB's cloud.

Essentials and Pro both work. Pro gets one to three read replicas, with automatic failover. If the primary stops responding, KiviDB promotes a replica and repoints the endpoint, and your application keeps the same hostname. Each database also gets a private name that resolves inside your network, so traffic from your own machines never leaves it.

Backups stay put

Disk snapshots and KDB exports are stored in your account, and they restore only into that same account and region. A backup can't be restored into KiviDB's cloud or into another of your connections by mistake. KiviDB enforces this in the API, so it holds for the console, Terraform and the MCP server alike.

You can also bring data in. In Snapshots, Import uploads a .kdb or .rdb file from your computer straight into your own bucket, ready to restore.

Terraform and AI assistants

The Terraform provider looks up a connected account by name and places a database in it:

data "kividb_cloud_account" "prod" {
  name = "Production GCP"
}

resource "kividb_instance" "orders" {
  name             = "orders"
  cloud            = "gcp"
  region           = "us-east1" # one of the account's regions
  tier             = "pro"
  data_size_gb     = 8
  replica_count    = 1
  cloud_account_id = data.kividb_cloud_account.prod.id
}

The MCP server does the same from an AI assistant: list_cloud_accounts shows your connections, and create_database takes a cloud_account_id.

What's different, for now

  • Each database needs a public endpoint, limited to its allowlist. Private-only databases are not available yet.
  • Free tier databases always run in KiviDB's cloud.
  • There's no private connectivity add-on, and you don't need one: the database is already inside your network.

Get started