Skip to content

How-to: target a single command at a named account without switching

This is a focused task covering the per-call account override. The full account model — storing accounts, switching the active one, and S3 keys — is in Manage multiple RunPod accounts and switching between them per command. For every flag and the precise resolution order, see rp doc auth.

Goal: run one rp command under a specific stored account while leaving the active account exactly as it was.

Steps

  1. Confirm the stored account names, then pass the global --account flag to any command:

$ rp auth list $ rp volume list --account personal

  1. Or set the RP_ACCOUNT environment variable, scoped to the single command via a leading assignment:

$ RP_ACCOUNT=personal rp volume list

  1. For scripts and CI, an exported RUNPOD_API_KEY (or RUNPOD_API_KEY_FILE, pointing at a file that holds the token — e.g. a mounted secret) beats every stored source, --account included:

$ RUNPOD_API_KEY=<key> rp volume list

Notes

  • Resolution order (highest first): an explicit exported RUNPOD_API_KEY/RUNPOD_API_KEY_FILE → --account/RP_ACCOUNT → the active pointer → the credentials.d/default file → the single account present (if exactly one) → the user/install .env → none.
  • --account/RP_ACCOUNT selects an account for the one call only; the active pointer is never modified. Switching the active account for all future calls is rp auth switch <name>.
  • Within the --account/RP_ACCOUNT tier, the RP_ACCOUNT environment variable takes precedence over the --account flag if both are set.
  • Exit codes are stable for scripting: transport auth failures (HTTP 401/403) exit 3, so a CI step can distinguish a credentials problem from a request problem (exit 1) or a missing resource (exit 4).