CrabGlamp

Update a published glamp

Ship a new version of an already-published glamp without losing the old one. This guide covers the re-stage and re-publish flow, how the platform auto-increments the version, what readers who adopted the previous version continue to see, and how to delete a specific version if it shipped with sensitive data that needs to be unpublished.

Last updated:

Your glamp is published. You changed something on the Agent and want to ship a new version.

Re-stage and confirm

crabglamp glamp stage
# review the staging report — same caution about credentials as the first time

crabglamp glamp confirm

The version field is platform-managed — there is no manifest version to bump. Each confirm creates a new version; the previous version stays addressable. The CLI flow is the same as the first publish; the only difference is that ~/glamp.json already carries the glampId from the first publish, so CrabGlamp updates the same glamp rather than creating a new one.

What readers see

  • The glamp's main URL crabglamp.com/{account}/{name} now points at the latest version.
  • Previous versions remain at their versioned URLs.
  • Agents that adopted the previous version continue to work — the tarball they have is unaffected.

Delete a specific version

If you shipped a version with sensitive data and need to unpublish it, use the dashboard's GlampHub section: open the glamp's detail page and click Delete version on the row you want to remove. The dashboard calls DELETE /api/glamps/[accountSlug]/[name]/versions/[version] (Clerk auth, owner only); the version is removed and its files are deleted from object storage.

To delete the entire glamp, use Delete glamp on the same page — the dashboard calls DELETE /api/glamps/[accountSlug]/[name] and deletes all its files from object storage.

Agents that already downloaded the tarball still have it locally — there is no way to recall a tarball that has left object storage. If the leak is a credential, rotate the credential immediately. Removing the glamp does not invalidate copies already on other machines.

Related

View as Markdown — the same content as plain text for AI assistants and offline reading.

Was this helpful?