> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nerdit.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Deploy your first application

> Install the Engine, deploy a small CPU application, and verify its local endpoint.

This walkthrough runs on the **Engine host**. It needs Docker and an internet connection to download the example image. No GPU or App account is needed.

<Steps>
  <Step title="Install and check the Engine">
    Follow [Installation](/engine/installation) for your operating system, including Docker permissions and the correct installation method. Return when these commands work:

    ```bash theme={null}
    nerdit --version
    nerdit doctor
    nerdit capabilities
    ```

    Use `nerdit doctor` to check Engine connectivity, Docker and the selected runtime. Resolve failures before deploying. The local dashboard, when bundled, is at `http://localhost:9321/`.

    <Warning>
      Do not run `nerdit init` on an installation already configured by the signed installer. Source installations have their own bootstrap steps in the installation guide.
    </Warning>
  </Step>

  <Step title="Create an example application">
    In an empty working directory:

    ```bash theme={null}
    mkdir nerdit-hello
    cd nerdit-hello
    cat > index.html <<'EOF'
    <h1>Hello from Nerdit</h1>
    EOF
    cat > Dockerfile <<'EOF'
    FROM python:3.13-alpine
    WORKDIR /app
    COPY index.html .
    USER 65532:65532
    EXPOSE 8000
    CMD ["python", "-m", "http.server", "8000", "--bind", "0.0.0.0"]
    EOF
    cat > nerdit.toml <<'EOF'
    [deploy]
    name = "nerdit-hello"
    port = 8000
    health = "/"
    EOF
    ```

    This uses Python's standard-library HTTP server as a disposable smoke test. Use your application's production server for a real deployment. The image listens on the declared port and runs without root privileges.
  </Step>

  <Step title="Validate and deploy">
    ```bash theme={null}
    nerdit deploy . --dry-run
    nerdit deploy . --wait --timeout 300
    nerdit services list
    ```

    The dry run validates the deployment and prints a plan without building an image or creating a service. The real deployment uploads this directory, builds the Dockerfile on the Engine host, starts a container and waits for health. The expected result is a converged deployment and a running service.

    A timeout exits with code `3`; the deployment can still be running. Continue waiting with `nerdit services wait nerdit-hello --timeout 300`, or inspect `nerdit diagnose nerdit-hello` and `nerdit logs nerdit-hello`.
  </Step>

  <Step title="Open the application">
    By default the Caddy proxy is disabled. Find the service's loopback endpoint in `nerdit services list` and open it **on the Engine host**. For example, if the allocated port is `9400`:

    ```bash theme={null}
    curl --fail http://127.0.0.1:9400/
    ```

    Use the actual port from your service, not an assumed port. The response should contain `Hello from Nerdit`. Loopback endpoints are not remotely reachable.

    For an HTTPS endpoint, follow [Networking and HTTPS](/engine/networking) to enable Caddy and select a hostname and listener port, then restart the Engine using the method that manages your installation. Check `nerdit proxy status` and `nerdit routes`. In path mode, open the listed URL including `/nerdit-hello/`.

    Trust the node's internal CA on each client after verifying its fingerprint through a trusted channel; see [Networking and HTTPS](/engine/networking). An internal certificate is not publicly trusted. A public domain and public certificate require the separate DNS and ACME steps in that guide.
  </Step>

  <Step title="Inspect and remove the example">
    ```bash theme={null}
    nerdit logs nerdit-hello
    nerdit diagnose nerdit-hello
    nerdit services stats nerdit-hello
    nerdit services stop nerdit-hello
    nerdit services rm nerdit-hello --purge secrets,images
    ```

    Removal deletes the service and the selected resources. Keep the source directory if you want to deploy it again. For applications with persistent data, read [Backups and recovery](/engine/backup-recovery) before choosing any data purge.
  </Step>
</Steps>

Continue with [Deploy applications](/engine/deploy), add [AI services](/engine/ai) or [managed databases](/engine/databases), or [connect the App](/app/connect-engine) to operate the same node from your account.