When you hire an agent without naming an environment, it now comes with its own computer that is already booting, so it is runnable the moment it exists.
A guided build. A guided flow walks you through the platform the computer runs, the repositories it clones, the services that boot with it, and the setup steps that install what it needs, then shows the exact memory and CPUs the machine gets on the platform you picked. The YAML manifest is still there when you want to edit it directly.
Tasks follow the agent. Starting a task begins with picking the agent, and the work runs on that agent's computer without you choosing one each time. Naming a different computer still wins.
Computers without repositories. A computer that sources no repositories generates its snapshot without a git token, so a research or ops agent runs with no GitHub connection. Add a repository later and the agent needs a GitHub identity from then on.
One machine, one name. The app says computer on every screen, and so do its setup emails and error messages. The API resource is still /v1/environments, the /accounts/:id/environments route still resolves, and the CLI still says environment, because renaming those would break published links and the scripts written against them.
Start with the agent you already have
Open the agent, go to Configure, and select Computer. If you hired it without naming one, a computer is already there, generating its snapshot. View computer opens it: Settings holds its name and description, and Build holds its manifest and the log from its last build. Build a new computer, in the same list, opens the guided flow.