# Install Redbelt and Get a First Result

Complete [Before Installing Redbelt](/self-hosting/before-install/). This procedure changes the target host and starts workloads. Plan downtime and use only the release approved for your deployment.

## Install from customer configuration

1. Obtain and verify the signed operator distribution and private customer configuration from Redbelt. Do not clone the product repository as an installation step.
2. Place the configuration at a protected path, such as `/etc/redbelt/customer.json`, readable only by the administrator who performs installation. Use the administrator privileges required by the supplied release instructions.
3. Run the operator with the intended private state directory:

```sh
redbeltctl --state-dir /var/lib/redbelt setup --config /etc/redbelt/customer.json
```

4. Enter the installation key at the hidden prompt. Automation can use `--key-stdin` with a protected secret-input mechanism; never place the key literally in a shell command.
5. Keep the operator's private state directory and recovery material. Enrollment records an operation before contacting the control plane. Setup resolves an authenticated, signed release and verifies downloaded artifacts before applying them. It does not authorize arbitrary images or caller-supplied download URLs.
6. Check progress using a second terminal:

```sh
redbeltctl --state-dir /var/lib/redbelt status
```

The installation key is a short-lived enrollment credential, distinct from a product API key. Its issuance includes an expiry; obtain a replacement through the issuing administrator if it expires before enrollment. A saved verified receipt permits recovery of the same enrollment intent but does not bypass current licensing or revocation checks.

## Interrupted installation

After fixing a reported prerequisite or connection failure, rerun the same setup command with the same configuration and state directory. Preserve the journal and completed checkpoints. Do not delete state or change release/configuration to force a retry. An `inspection_required` outcome needs administrator review before further changes. See [recovery](/self-hosting/recovery/).

## First sign-in and first result

1. Open the application HTTPS origin supplied in your configuration.
2. Sign in as the designated deployment administrator. Enrollment alone does not sign you in or grant another account administrator access.
3. Complete the offered model-provider setup. Enter your authorized provider credentials and run the offered tests; a model list alone does not prove inference works.
4. Create or select a workspace and [start a thread](/threads/first/). Send a short request and check the completed answer.
5. Check the deployment's documented live/readiness endpoints with your administrator. A saved configuration or `installed` status is not a full application health check.

If sign-in fails, confirm the exact application origin and administrator identity rather than changing identity-provider settings blindly. If the first request fails, inspect the model test, budget and error. Successful installation does not automatically connect third-party apps.
