Harbor adds Docker Sandboxes to run AI agent evals in isolated microVMs

Docker announced that Harbor, its open-source agent evaluation framework, now runs each task in its own Docker Sandbox microVM with a separate kernel. Docker tested 64 concurrent tasks on Cloud Sandboxes, with a stable release due October 23.

• Each Harbor task runs in its own microVM with a separate kernel
• Docker tested 64 concurrent tasks on Cloud Sandboxes
• Network policies cut internet access after dependency setup
• Available in nightly builds, stable release on October 23