Cloud Computer AI provides isolated cloud desktops for intelligent agents. This guide outlines the concepts and a practical path from first credentials to a controllable machine.
A Cloud Computer AI computer is a full graphical environment: a desktop session with a browser, filesystem, and terminal. Your software creates a machine, receives a stream or control endpoints, and drives the session through screenshots, pointer events, keystrokes, and optional shell commands. Cloud Computer AI is deliberately not the agent itself; it is the computer on which the agent works.
POST /api/computers in the current API surface).The visual console supports infinite-canvas workspaces, templates, invitations, comments, and operational monitoring of sandboxes. Prefer the console when onboarding teammates or supervising sessions in real time.
Programmatic access covers computer lifecycle, streaming, remote execution, and operational listings used by the E3B monitoring views. Authenticate with the credentials issued for your workspace; never embed secrets in client-side code intended for the public internet.
Templates define the starting state of a machine. Cloud Computer AI ships curated options for popular agent harnesses and a blank desktop for custom stacks. Selection of a template should result in a machine prepared accordingly—not merely a label—so that agents begin from a predictable environment.
Detailed reference pages will grow as the public API stabilizes. Until then, the console and health endpoints remain the most reliable way to explore the product.