AI agents can run commands, modify files, and download files from untrusted sources directly on a developer machine, which creates a major security risk. There needs to be a way to safely run agents and isolate how it interact with the network, files, host system, etc.
Docker Sandboxes solves this problem by creating isolated microVM environments where AI agents run safely with all the guardrails without affecting the host system. Docker Sandboxes support Claude Code, Codex, Cursor, etc. A complete list of agents can be found here.
Prerequisites
macOS Sonoma (version 14) or later
Apple silicon
