Your workspace and data
Understand what Claude has in a pod, what survives a restart, and what still needs a backup.
A pod is a private, project-scoped cloud computer. It gives Claude enough of a real machine to build and operate useful software without giving it your entire laptop.
What Claude can keep together
- The project: source code, working notes, configuration, and repository history.
- The toolchain: project packages and the language or framework tools supplied by the chosen environment.
- Supporting software: development servers, workers, Postgres, and other local services.
- Working data: development databases, generated reports, and files stored in the persistent workspace.
- The agent's context: project instructions, skills, and the Claude session attached to the pod.
Keeping these parts in one place is what lets Claude verify a complete result instead of handing you code that still needs to be assembled elsewhere.
What survives
Files in the pod's home workspace survive a restart, suspension, and image update. Claude keeps the repository and important local data there.
Running processes do not survive by themselves. Tell Claude which application, worker, or helper must come back automatically; Podbay can supervise those services and relaunch them after a restart. Scheduled work is stored durably as well, but it only runs while the pod is on.
Temporary files and changes to the underlying operating-system image may disappear after an update. Claude has live information about these boundaries and should choose a durable project-level setup when the result needs to last.
What to put in Git or another backup
The persistent workspace is convenient state, not an independent backup. Commit and push valuable source changes. Export databases, uploaded files, generated reports, or other data that would be costly to recreate.
Deleting a pod removes its persistent workspace. Suspension does not.
A pod is not production hosting by default
A pod is excellent for development, prototypes, internal tools, and small always-on workflows. Use independently backed-up services when the application needs production availability, compliance, or data guarantees beyond one project machine.