Skip to content

You built it. Put it somewhere real.

A local project becomes much more useful when someone else can open it. Bring what you have built, give your agent a deployment target, and keep improving the same app after it is live.

Start deployingRead the docs

Start with your own project or a useful template

Use the code already on your machine or in GitHub. If the blank page is slowing you down, explore business app concepts and landing page templates. Their prompts, previews, and source files can make the starting point concrete.

Choose a template for the task rather than its thumbnail alone. Read what is included, which interactions are local demonstrations, and which services need to be connected for your use case.

Get the essentials working first

Test the core inputs, outputs, and error states before deployment. If the tool calculates a result, verify the calculation. If it exports a file, open the file. If it submits to another service, confirm that the real service received the expected request.

Use the smallest architecture that fulfills the job. A simple static utility may not need a backend, while shared records or private data require a deliberate data and authentication design.

Keep the project moving after launch

Install the build.host skill, sign in, and ask your agent to deploy the existing folder or repository. Use one recognizable project name. After a successful deployment, open the live URL and repeat the workflow you tested locally.

When you revise the tool, update that same project. Your agent can inspect logs, configure variables, redeploy, or restore an available earlier version. Approved capabilities and external service setup still govern what the app can run.

Good questions.

Do I need GitHub to start?

No. You can deploy a local folder directly. GitHub is available when you want a repository-based source workflow.

Are business templates already complete production systems?

Check the entry’s label and included files. The current business collection provides visual concepts and complete implementation prompts; production behavior and integrations need to be built and verified.