Show HN: Ramet – Persistent, forkable workspaces for AI agents (github.com)

If I switch from one sandbox to another, this keeps the agent context in tact? A lot of sandboxes offer audit trails so is it more than simply copying and pasting that context?