AI agent glossary
Sandboxing
Running an agent inside an isolated environment, such as a container or virtual machine, so any mistakes or attacks stay contained.
A sandbox limits what an agent's actions can reach. Inside it, the agent can run commands, install packages, and edit files freely. Outside it, your real files, credentials, and network are protected.
Common sandboxes include containers, virtual machines, and restricted operating system profiles that block file access outside a project folder or limit network connections. Cloud agent platforms usually run each task in its own disposable environment.
Sandboxing is what makes higher autonomy safe enough to use. When the worst an agent can do is break a throwaway environment, you can let it work without approving every step.
A coding agent runs in a container with only the project folder mounted and network access limited to the package registry, so it cannot read your SSH keys or upload files elsewhere.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 65 practical lessons.