Skip to content
These docs are a preview of Self-Managed Kindo 2026.11, an upcoming release.

How the config works

Every config setting lives in one config stored in the cluster, so anyone with cluster access and the Kindo CLI can change the instance. Every change follows the same steps: change the value, then apply it. The apply shows what it will change and asks before it does.

Change Kindo only through the CLI. Applying the config puts back what was changed directly in the cluster, such as an environment variable or a ConfigMap value, and the plan lists each object it puts back, without Secret values. Only an upgrade changes the installed version. Three cases differ:

  • Images: one changed by hand to the installed version or an older one is put back. A newer one stops the apply. An image the CLI can’t check, such as a third-party one, is only listed.

  • Replica counts: a count an autoscaler manages is left alone. A count set by hand stays until the component is deployed again, when it returns to its chart’s count. To keep a count, set the chart’s replica value under the component’s name in the editor:

    overrides:
    helmValues:
    <component>:
    replicaCount: 3
  • Not checked: the plan notes any group it can’t compare, such as the integrations. Changes made by hand there, other than images, aren’t listed and stay unless their component is deployed again for another reason.

The Config reference describes every config setting.

  1. Change a single value:

    Terminal window
    kindo config set inputs.smtp.host=smtp2.example.com
    kindo config set inputs.smtp.auth.login.password # prompts for the secret
    kindo config unset inputs.smtp.port # back to its default

    For a whole section, a list, or a Helm value, edit the config in your editor instead:

    Terminal window
    kindo config edit

    Every change is checked before it’s saved, and nothing changes in the cluster until you apply it.

  2. Read back what you changed. Credentials stay hidden unless you ask to see them.

    Terminal window
    kindo config get inputs.smtp
    kindo config get inputs.smtp --show-secrets
  3. Apply the change:

    Terminal window
    kindo config apply

    Applying updates the components whose configuration changed, at the version that’s installed, and restarts their workloads. Commands lists everything it does.

You can narrow an apply to some components, or restart one of Kindo’s applications whose configuration didn’t change:

ToRun
Apply only some componentskindo config apply --only api,task-worker-ts
Re-run only the integrationskindo config apply --only integrations
Restart an application’s podskindo config apply --restart --only <component>

To move to another version, upgrade.

Config settings you can’t change after installing

Section titled “Config settings you can’t change after installing”

Once Kindo is installed, you can’t change the storage class, because its data volumes keep the class they were created with. A change to another class is refused, and so is applying a config that names one. The refusal names the class the volumes use.

You also can’t move a database to another PostgreSQL server. Applying the config, upgrading and moving to another secret store all refuse a config that puts any database on a different host or port, including one that adds or removes the auxiliary or Hatchet server, and list each database that would move. A new password, login or TLS config setting on the same server applies as usual.

To keep the instance, set the servers back to the ones its databases are on:

Terminal window
kindo config edit

To use different servers, install a new instance with them.

Change the config with the installed version’s CLI; an older CLI refuses. A newer CLI can change it too, but the change takes effect when you upgrade to that version, because applying refuses a newer CLI. Until installing finishes, only the CLI version that started it can change the config.

KindWhere it livesExample
Config settingsinputsinputs.smtp.host
Environment overridesoverrides.serviceEnv.<app>overrides.serviceEnv.api.LOG_LEVEL=debug
Helm value overridesoverrides.helmValues.<component>, in the editoroverrides.helmValues.api.resources.requests.memory: 4Gi

Use a config setting whenever one exists. The CLI refuses a credential set as an environment override and says what to do instead. Overrides stay in place across upgrades, so you set each one once. An entry with the same name in your secrets manager takes precedence over an environment override, so remove it from your secrets manager for the override to take effect.

For example, to raise one application’s log level:

Terminal window
kindo config set overrides.serviceEnv.api.LOG_LEVEL=debug

For LiteLLM, use its own variable:

Terminal window
kindo config set overrides.serviceEnv.litellm.LITELLM_LOG=debug

To override a Helm value, add it under the component’s name in the editor:

overrides:
helmValues:
api:
resources.requests.memory: 4Gi
resources.limits.memory: 4Gi

A value keeps the type you write. Leave true, false and numbers unquoted, and quote a value that must stay text, such as an annotation or an environment variable:

overrides:
helmValues:
qdrant:
replicaCount: 3
litellm:
envData.LITELLM_LOG: "DEBUG"

Kindo runs Qdrant, for vector search, and Speaches, for speech-to-text, in the cluster. To use a Qdrant you run, or to turn off Speaches, change them in the editor:

Terminal window
kindo config edit
inputs:
qdrant:
bringYourOwn:
url: https://qdrant.example.com:6333
apiKey: <your-qdrant-api-key>
speaches:
none: {}

Apply the change:

Terminal window
kindo config apply

To have Kindo run either one again, replace its config setting with:

bundled: {}

After you change a value in your secrets manager, wait for it to sync into the cluster. Then restart the applications that read it:

Terminal window
kindo config apply --restart --only <component>