Commands
Each command’s help lists all its options. This page covers what each command does, what it needs before it runs, and what it refuses.
How every command behaves
Section titled “How every command behaves”-
Commands act on your current Kubernetes context, the same way kubectl and Helm do, unless you name another context. Every plan names the cluster it will change.
-
Upgrades change the version; applies change everything else.
-
A command that changes the cluster shows its plan and asks first. That covers install, upgrade, restore, uninstall, resetting a database, the health check’s repair, applying the config, the config migration, moving the secret store, copying the feature flags and resetting records of earlier runs. The plan reports the problems it can find in advance, and the command stops when it finds one. Answering no changes nothing. The infrastructure checks, including the test email, run only after you answer yes.
-
The infrastructure checks stop a run only when one fails. Install, upgrade, restore and applying the config run them before deploying anything. A failed check stops the run before it changes anything. A check that only warns is listed with its fix, and the run continues.
-
A command that changes the cluster can show only its plan, or run without asking. A script or CI job has to run without asking: without a terminal to ask on, the command stops before it changes anything.
To Add Show only the plan --planRun without asking --apply -
Reading commands change nothing. These leave the instance as it was:
- status
- the logins of the supporting services
- the health check
- reading the config
- taking a backup
- every plan
- every listing
-
A command that can’t finish stops and fails. It says what blocked it, and exits non-zero.
-
Output is plain when it isn’t going to a terminal, such as in a script, a CI job or a log file: no colour or animation, and no wrapped lines. In a terminal, you can ask for plain output:
Terminal window kindo --plain <command>KINDO_PLAIN=1 kindo <command> -
A workstation with little memory can limit how much runs at once. See the memory limit in Preparation for some instances.
-
One command changes the cluster at a time. A command that finds another one running stops. When nothing is running, it can be tried again after 10 minutes, or the message explains how to take over.
-
A 2026.09 instance needs the config migration first. Every command that reads the config refuses a 2026.09 instance until you migrate its config. See the config migration.
kindo config init
Section titled “kindo config init”Saves a new instance’s config to the cluster, from a config template, or by asking you questions. It generates the instance’s passwords and encryption keys. It can also print an annotated config template.
kindo config init --template > config.yamlkindo config init --file config.yamlkindo config initNeeds: a cluster with no Kindo config and no Kindo instance.
| Refuses when | Points to |
|---|---|
| The cluster already has a config | kindo config edit |
| Kindo is installed, but the config is missing | kindo doctor |
| A 2026.09 instance that hasn’t been migrated | kindo config migrate |
| A config template that still has values in angle brackets | replacing each one, as the message lists them, with your own |
kindo install
Section titled “kindo install”Deploys the Kindo CLI’s version from the config, or finishes installing after a run that stopped partway. It checks your infrastructure before deploying anything, including a test email, and creates the first superadmin from the config’s superadmin email.
kindo installkindo install --no-test-emailNeeds:
- A config with a superadmin email
- The tools on your workstation (Prepare your workstation)
- Infrastructure checks that pass
- No instance that finished installing
| Refuses when | Points to |
|---|---|
| The cluster has no config | kindo config init |
| Kindo is already installed | kindo upgrade for a new version; kindo config apply for a config change |
A kindo install run that another version started and didn’t finish | that version’s CLI |
| A pending release it can’t roll back by itself | the command to clear it, then kindo install |
| A pending release you turned off in the config | the Helm command it names, then kindo install |
| Your external secrets manager lacks a key installing needs | the missing keys |
Your external secrets manager’s login Secret is missing from kindo-system, or lacks a key | recreating that Secret, which the message names with its keys |
| The registry doesn’t answer a chart lookup | running it again once the registry recovers |
kindo upgrade
Section titled “kindo upgrade”Moves an instance that finished installing to the CLI’s version.
kindo upgradekindo upgrade --no-test-emailNeeds:
- An instance that finished installing, on 2026.09 (after the config migration) or on the CLI’s own release
- The version’s charts in the registry
- A healthy cluster
- Infrastructure checks that pass, including a test email unless you skip it
| Refuses when | Points to |
|---|---|
| No instance, or one that didn’t finish installing | kindo install |
| The version that’s already installed, unless an upgrade is unfinished or a command told you to run it | kindo config apply |
| A missing or incomplete config | kindo uninstall all --keep-data, then kindo restore |
| A damaged cluster | kindo doctor |
| Records of earlier runs the Kindo CLI can’t read | kindo install-state reset |
| A pending release Kindo didn’t create, or one you turned off in the config | the Helm command it names |
| An instance on a newer release | rolling back a release line |
| An instance on a version before 2026.09 | the latest 2026.09 patch, with the 2026.09 CLI |
| A change to the public URLs Kindo’s applications serve | the config settings that set them |
| A config that moves a database to another PostgreSQL server | kindo config edit to set the servers back, or a new instance |
| The registry doesn’t answer a chart lookup | running it again once the registry recovers |
kindo config edit, set, unset, get
Section titled “kindo config edit, set, unset, get”Read and change the config stored in the cluster. Every change is checked before it’s saved, and nothing reaches the running instance until you apply it.
| Command | Does |
|---|---|
kindo config edit | Opens the config in your editor |
kindo config set <path>=<value> | Changes one value; a credential’s path alone prompts for it |
kindo config unset <path> | Returns one value to its default, or removes an override |
kindo config get [<path>] | Prints values, with credentials hidden |
kindo config get [<path>] --show-secrets | Prints values, with credentials shown |
Needs:
- A config that nobody else changed since the command read it
- To change it, the installed version’s CLI or a newer one, whose change takes effect at the upgrade. Until installing finishes, only the CLI version that started it.
| Refuses when | Points to |
|---|---|
| An invalid value | the config setting that’s wrong |
| Setting a whole section, a list or a Helm value at once | kindo config edit |
| A credential set as an environment override | the config setting for that credential |
| A change of secret store once installing has started | kindo secret-store migrate |
| No config, or an empty one | kindo config init for a new instance; kindo restore for a lost one |
| A change with a CLI older than the installed version, or before installing finishes, with a CLI other than the one that started it | the CLI version it needs |
The passwords and keys the CLI generates aren’t shown in the editor and can’t be changed.
kindo config apply
Section titled “kindo config apply”Applies the config at the installed version. It updates and restarts what changed, deploys the components and integrations you turned on, and, unless you narrowed the run, adds the version’s feature flags your instance doesn’t have yet. It rebuilds each database a reset left empty. It also puts back what was changed directly in the cluster, including a registry login added by hand; the plan lists each changed object. See How the config works.
kindo config applyTo narrow an apply to some components or restart an application, see How the config works.
Needs:
- An instance that finished installing
- The CLI and the instance on the same version
- A complete config
- A healthy cluster
- Infrastructure checks that pass
| Refuses when | Points to |
|---|---|
| No instance, or one that didn’t finish installing | kindo install |
| The CLI and the instance on different versions | the installed version’s CLI; with a newer CLI, kindo upgrade |
| An unfinished upgrade, unless only its integrations or feature flags are left | kindo upgrade, to finish the upgrade |
| A missing config, or lost encryption keys | kindo uninstall all --keep-data, then kindo restore |
| Generated passwords or keys the config lacks | kindo upgrade |
| A damaged cluster | kindo doctor |
| Records of earlier runs the Kindo CLI can’t read | kindo install-state reset |
| A pending release Kindo didn’t create, or one you turned off in the config | the Helm command it names |
| Config settings it can’t apply | kindo config edit |
| A config that moves a database to another PostgreSQL server | kindo config edit to set the servers back, or a new instance |
| A Kindo image changed by hand to a newer version | finishing that upgrade with that version’s CLI, or rolling back from a backup taken before it |
With --only integrations, integrations it can’t plan | the cause the message names |
With --only, a run that leaves out a component a database reset left to rebuild, or any run after a Nango reset | kindo config apply without --only |
A component you turn off stays deployed. The plan shows the command that removes it.
kindo config migrate
Section titled “kindo config migrate”The one-time config migration of a 2026.09 instance’s config settings into the config. It reads the cluster’s 2026.09 configuration and, when it finds them, your install contract and environment bindings files, and asks you about each config setting where they disagree. See Upgrade from 2026.09 to 2026.11.
kindo config migratekindo config migrate --apply --resolve <setting>=<answer>Needs:
- A 2026.09 instance that finished installing, with every component it runs on the same 2026.09 patch and no pending releases or components being removed
- Every application’s credentials delivered through the External Secrets Operator
- An answer to every question
If the new config would change your running instance, it stops without saving.
| Refuses when | Points to |
|---|---|
| An instance not on 2026.09 | upgrading to 2026.09 with the 2026.09 CLI |
A kindo install or kindo upgrade run on 2026.09 that didn’t finish | finishing it with the 2026.09 CLI |
| An application on plain Secret delivery | the latest 2026.09 patch, with the 2026.09 CLI, after turning back on any component it names |
| Components on more than one 2026.09 patch | the latest 2026.09 patch, with the 2026.09 CLI |
| A turned-off component still on a version before 2026.09 | turning it back on and upgrading with the 2026.09 CLI |
| A pending release | letting it finish, or the 2026.09 CLI’s kindo doctor --fix |
| A disagreement it can’t resolve unattended | an answer for that config setting on the command line |
kindo secret-store migrate
Section titled “kindo secret-store migrate”Moves an instance from Kubernetes Secrets to Vault, AWS Secrets Manager or Azure Key Vault. Your secrets manager must hold the entries and values in Secret store keys. The move checks that it does, saves the new store config setting, and updates the components. See Secret store.
kindo secret-store migrate --to <vault|awsSecretsManager|azureKeyVault> --settings <file> --plankindo secret-store migrate --to <vault|awsSecretsManager|azureKeyVault> --settings <file>Needs: everything applying the config needs; your secrets manager holding every listed key.
| Refuses when | Points to |
|---|---|
| Any move other than from Kubernetes Secrets | not supported |
| A Helm override that changes how an application reads its credentials | kindo config edit |
| A config that moves a database to another PostgreSQL server | kindo config edit to set the servers back, or a new instance |
| The CLI and the instance on different versions, or an unfinished upgrade | the installed version’s CLI, or kindo upgrade |
| A damaged cluster, or an instance whose state can’t be read | kindo doctor |
| Records of earlier runs the Kindo CLI can’t read | kindo install-state reset |
If it stops after saving the new store config setting, applying the config finishes it.
kindo config backup
Section titled “kindo config backup”Writes an encrypted backup of the config and the instance’s state to a local file. It asks for a passphrase, or reads it from an environment variable. In a script or CI job the variable is required: without a terminal to ask, the backup stops and names it. Restoring an encrypted backup works the same way.
kindo config backup --output <file>KINDO_BACKUP_PASSPHRASE=<passphrase> kindo config backup --output <file>kindo config backup --output <file> --allow-incompleteNeeds: an instance on this CLI’s release, and a complete config. The last form backs up a config that’s missing values that can’t be generated again, and that backup can’t fully restore the instance.
| Refuses when | Points to |
|---|---|
| No config, or an empty one | kindo restore |
| An instance on an earlier release that hasn’t been migrated | a backup with that release’s CLI |
| An instance that has been migrated but not yet upgraded | kindo upgrade, then the backup |
| A config missing only Hatchet’s worker token, as after a reset of Hatchet’s database or a run that stopped | the command that has Hatchet create one: kindo install if installing or restoring stopped, kindo upgrade with the CLI the refusal names if an upgrade stopped, otherwise kindo config apply; then the backup |
| A config missing other values that can’t be generated again, such as encryption keys | kindo uninstall all --keep-data, then kindo restore with a backup taken before they were lost; or the last form, which backs up without them |
kindo restore
Section titled “kindo restore”Installs Kindo from a backup onto a cluster with no Kindo instance, at the CLI’s version. It shows what it will do and asks before starting. You can change your own config settings on the way in, except the secret store and integrations.
kindo restore <file>kindo restore <file> --set inputs.postgres.main.host=<host>kindo restore <file> --apply # don't ask before startingNeeds: no Kindo instance on the cluster; a complete backup from any patch of this CLI’s release.
| Refuses when | Points to |
|---|---|
| Kindo is installed on the cluster | kindo uninstall all --keep-data, then the restore |
| A backup from another release | that release’s CLI |
Your external secrets manager’s login Secret is missing from kindo-system, or lacks a key | recreating that Secret, which the message names with its keys |
| Signing in to the registry fails | running the restore again with --set inputs.registry.username=<username> --set inputs.registry.password=<password> |
A value the backup doesn’t hold can’t be restored.
kindo uninstall
Section titled “kindo uninstall”Removes what the CLI installed, data volumes included: everything, or only the applications or the supporting services.
kindo uninstall allkindo uninstall applicationskindo uninstall peripheries| To | Add |
|---|---|
| Keep the data volumes and the sandbox workspaces. With no usable config, the components stay deployed too | --keep-data |
Also remove the namespaces, except those kept for components you run yourself, the config and its keys, the External Secrets Operator the CLI installed, the sandbox CRDs, and any login Secret for your secrets manager you created in kindo-system | --delete-namespaces |
Remove the finalizers of volumes or namespaces still being deleted after the wait, so they’re deleted without their owner’s cleanup. Not with --keep-data | --force-finalizers |
| Uninstall without asking (needed in a script or CI job, where nothing can answer) | --apply |
You can’t keep the data and remove the namespaces in one run, and removing the namespaces applies only to uninstalling everything. If your external secrets manager’s login Secret is in the kindo-system namespace, the uninstall warns before you confirm that removing the namespaces deletes it, and names the uninstall that keeps the namespaces and the commands that create the Secret again. Your external databases, your object storage, and an External Secrets Operator the CLI didn’t install stay as they are.
Components you run yourself or turned off
Section titled “Components you run yourself or turned off”When the config turns off Qdrant, RabbitMQ or Speaches, you may run your own, so the uninstall doesn’t remove it. When you also remove the namespaces, it keeps the namespace of any of these or of the S3 gateway. What it does with the volumes, and the namespace, depends on whether Helm lists Kindo’s release of the component there. A component the config turns on has its volumes deleted.
| Helm shows | Uninstall without --keep-data | Uninstall with --delete-namespaces |
|---|---|---|
| No release | Keeps the volumes and says why in the plan | Keeps the namespace and says why in the plan |
| The release | Keeps the volumes and names the commands that remove the release, then its volumes | Stops before changing anything, and names the command that removes the release |
| Can’t be listed | Keeps the volumes and names the listing to check | Stops before changing anything, and names the listing to check |
Uninstall can’t tell Kindo’s release from one you run yourself under the same name. For Kindo’s release, the commands it names remove it, and the uninstall is then run again. For your own, uninstalling with the data kept leaves it.
If the config is missing or can’t be read, the uninstall keeps those namespaces unless Helm lists the component’s release there.
kindo preflight
Section titled “kindo preflight”Checks the infrastructure against the config on its own, before installing or on an installed instance: databases, Redis, your own RabbitMQ, S3 object storage, email (including a test email to the superadmin, which you confirm arrived), the Kubernetes version, DNS, ingress, nodes and the default StorageClass. A check is skipped for Azure Blob Storage and for a RabbitMQ that Kindo runs. On an upgrade from 2026.09, the SSO sign-in accounts check refuses when an SSO sign-in is recorded more than once, and lists each record so you can delete the extras. It summarizes the results and lists each check that failed, warned or was skipped, with its fix or reason. It changes no application or config, and removes the resources its checks created when it finishes. The kindo-system namespace stays.
kindo preflightkindo preflight --no-test-email # skip the test emailkindo preflight --verbose # list every checkNeeds:
- A config
- The tools on your workstation (Prepare your workstation)
- A reachable cluster
It fails when a check fails or when it can’t remove something it created; the message names the item and the command that removes it.
| Refuses when | Points to |
|---|---|
The instance is on another version than the CLI, other than a 2026.09 instance migrated with kindo config migrate | that version’s CLI; kindo upgrade when the CLI is newer |
kindo doctor
Section titled “kindo doctor”Checks the instance and says how to fix each problem.
kindo doctorkindo doctor --fixkindo doctor --fix --apply # repair without askingNeeds: a reachable cluster with Kindo on it; for a repair, an instance that finished installing.
It fails when it finds a critical problem, can’t reach the cluster, finds no Kindo on the cluster, or a repair fails.
kindo install-state reset
Section titled “kindo install-state reset”Resets the records of earlier runs that the Kindo CLI can’t read, which the health check reports. Only those records change, and a step it resets runs again the next time its command runs.
kindo install-state reset| Refuses when | Points to |
|---|---|
| An instance on a newer version than the CLI | that version’s CLI |
| A 2026.09 instance that hasn’t been migrated | Kindo support |
| The records changed after the plan was shown | running it again to see the new plan |
kindo status
Section titled “kindo status”Reports the installed version and whether the instance is healthy, then lists what isn’t fine: steps that didn’t finish, components on the wrong version or unhealthy, and a missing superadmin. A step left running when installing stopped shows as interrupted, and status names the install command that continues it.
kindo statuskindo status --no-live # only what's recorded, without checking the clusterkindo status --verbose # every step and componentkindo validate
Section titled “kindo validate”Runs the end-to-end check against a running instance, as a test user in the system organization. It creates temporary API credentials and test data, temporarily changes the system organization’s DLP defaults, and removes the credentials and restores the defaults when it finishes. See Verify Kindo.
kindo validate --api-base-url <api-url> --base-url <app-url> --api-key-file <path>Needs:
- A working sandbox and models that can call tools
- A superadmin API key
- The configured address of the Kindo web app
It fails when a check fails or when it can’t remove its temporary credentials or restore the DLP defaults.
kindo integrations
Section titled “kindo integrations”Turns integrations on and off in the config, and lists them and their status. Applying the config deploys the change. Disabling one keeps its existing connections.
kindo integrations enable <name>kindo integrations disable <name>kindo integrations listkindo integrations statusEnabling and disabling refuse when the CLI and the instance are on different versions, and point to the installed version’s CLI, or to upgrading.
kindo feature-flags migrate
Section titled “kindo feature-flags migrate”Copies the instance’s feature flags from Unleash into Flagsmith. Once every copied flag matches, the next apply of the config switches Kindo’s applications to Flagsmith. Once the copy is done, the command refuses to copy again unless you ask it to overwrite Flagsmith, which replaces flags changed in Flagsmith.
kindo feature-flags migratekindo feature-flags migrate --overwrite-flagsmithNeeds: an instance that finished installing.
kindo logins
Section titled “kindo logins”Prints the address, username and password of the dashboards of the supporting services that ship with Kindo, such as HyperDX, Hatchet, Flagsmith and Nango. Passwords stay hidden unless you ask to show them. It changes nothing.
kindo loginskindo logins --show-secretsA supporting service that isn’t installed shows as not installed. When no ingress routes a service’s address, the command prints a port-forward for Flagsmith or Nango, and points to the ingress manifest for HyperDX and Hatchet.
kindo ingress manifest
Section titled “kindo ingress manifest”Lists every public hostname, the service behind it, its port and its health-check path, so you can set up DNS and your own ingress. Before you install Kindo it reads your config template; after that, the config.
kindo ingress manifest --file config.yamlkindo ingress manifestkindo ingress manifest --format json # or yamlkindo db
Section titled “kindo db”Connects you to Kindo’s databases through the cluster. See Databases.
| Command | Does |
|---|---|
kindo db list | Lists the databases |
kindo db tunnel <database> | Opens a tunnel to a database |
kindo db prompt <database> | Opens a psql session |
kindo db bootstrap-status | Checks that the databases were created |
kindo db reset <database> | Deletes everything in a database. The service doesn’t work until the next kindo config apply; see Reset refusals |
Reset refusals
Section titled “Reset refusals”The reset checks, before it changes anything, that it can read the keys Flagsmith or Hatchet issued.
| Refuses when | Points to |
|---|---|
| It can’t read the service’s keys (Flagsmith or Hatchet) | the cause its message names |
| Hatchet can’t be stopped first | the cause its message names; nothing was dropped |
If a reset stops partway, its message says to run the reset again, or that the next apply, install, restore or upgrade finishes forgetting the old keys before it deploys anything.
Until you apply the config after a reset, applying only some components and taking a backup refuse too; see kindo config apply and kindo config backup.
| Refuses when | Points to |
|---|---|
| The health check’s repair, after resetting Hatchet | the apply |
| Moving to your secrets manager, after resetting Hatchet | the apply |
Removed commands
Section titled “Removed commands”The release notes list the commands and options removed in each release line, and what to use instead.
