Why I Map AI Agent Workspaces to Drives Instead of Docker Volumes

Docker volumes work. But for Windows devs running AI coding agents, a mapped drive letter is simpler.

The difference

• Docker volume: needs a container running, path translation, networking, image management
• Drive letter: Windows sees a real drive, every tool works, zero overhead

My setup

• Local agent workspaces: subst.scsoi.com maps a folder to a drive letter
• Remote/WebDAV agent workspaces: drive.scsoi.com mounts the storage as a drive

The agent just sees a normal drive. It can write there freely. It cannot see anything outside that drive. Done.

When Docker still wins

You need dependency isolation, network isolation, or Linux tooling. Go Docker.

But if your only worry is "the agent will delete my files," a drive letter is enough. Its the cheapest sandbox you can build.