Deploy with an AI agent

One plain-language request, a handful of commands you approve, and you end with a real service running on your own cloud. It is the same result as Deploy your first application, except your AI coding agent runs the steps for you.

The hands-on part is a few minutes of describing what you want and approving each command. The clock is mostly the first environment launch, which provisions a real cloud environment (network, cluster, storage) and is the longest step, so budget roughly 15 to 30 minutes end to end on a first run, most of it spent waiting rather than typing.

By the end you will have a project on Facets, a service running the public nginx image, a dev environment launched on your connected cloud account, and a release that deployed the service and shows it Running with live logs.

Prerequisites

The agent drives your control plane through the Praxis CLI, so it needs the same one-time setup as the manual quickstart. Have these in place first:

• A Facets tenant you can sign in to (Facets hosted or Self-hosted).
• A cloud account and a git account connected. The cloud account is where the environment launches, without one there is nowhere to deploy, and the git account is where your project's blueprint is stored.
• The Praxis CLI installed and logged in. Signing in installs the Facets skills into your AI host (Claude Code, Gemini CLI, and others), so it knows the blueprint model, the module contract, and which commands to run.

Describe the outcome

Open your AI host and tell it what you want in plain language, for example:

Import a project type, create a project called checkout, add a service that runs the public nginx image, launch a dev environment, then release it so I have a running app.

With the Facets skills installed, the agent reaches for the right Praxis CLI commands without being told the syntax. It works through the same steps the manual quickstart covers:

• Importing a project type to populate a new control plane.
• Creating a project and adding a service to its blueprint, running the public nginx image.
• Launching an environment on your cloud account, then releasing your changes and reading the logs if a release fails.

You stay in control. The agent proposes each command and you approve it, and everything it runs goes through the Praxis CLI against the same control plane you would use by hand, so everything it does, you could audit or repeat yourself.

You now have a running application

When the release finishes, your dev environment shows as Running and your service is live on your own cloud, with its status, logs, and metrics in the Resource Center. You got there by describing the outcome once and approving the steps, and because every step ran through the Praxis CLI, you can repeat it, script it, or hand the next change to the agent again.

To go further:

• Deploy your first application for the manual, step-by-step path to the same result.
• Core concepts for the model the agent is working with.