Docker

Docker now offers Cloud Sandboxes, letting developers run AI agent workflows in Docker-managed cloud infrastructure with the same microVM isolation, CLI, trust model, and Kits used in local sandboxes, so agents keep running after a laptop shuts down. This is a new solution rather than a breaking change, so no existing setup requires modification; developers can extend local sandbox configurations to the cloud or start new ones without switching environments.

Matters for: Teams building and deploying long-running, unattended AI agent workflows that need to scale beyond a single developer's laptop while maintaining consistent isolation policies.

Docker now offers Cloud Sandboxes, the same microVM-based isolation model as local Docker Sandboxes but running on Docker-managed compute, letting agents run unattended for hours and scale to many parallel tasks; a single CLI command (sbx move) moves a sandbox's filesystem between local and cloud in either direction. Cloud Sandboxes ship with pre-configured kits for agents like Claude Code, Codex, and Copilot, an MCP gateway for connecting tools such as Jira and Grafana, secret injection that hides actual secret values from agents, and configurable network policies.

Matters for: Teams building autonomous coding agents that need to run long-horizon tasks (multi-hour refactors, dependency migrations, test suites) unattended and in parallel without provisioning their own cloud infrastructure.