Skip to main content
Run promptster-teams <command> [flags]. In normal use you only run login once and occasionally status or doctor — the rest manage background capture or help you recover.

Capture

Saves your developer key, validates it, pings your dashboard host, and starts background capture scoped to your home directory. This is the only command most engineers run.
See Connect for the full flow.
Spawns a detached capture process and returns your shell. Idempotent — running it when capture is already active does nothing. Logs go to ~/.promptster-teams/daemon.log.
login already calls this for you; use start directly if you connected via the PROMPTSTER_TEAMS_TOKEN environment variable.
Stops the background capture process.
It signals the tracked process IDs recorded under ~/.promptster-teams/. If capture was started without a tracked PID, stop tells you to find it with pgrep -fl promptster-teams.
Runs capture in the foreground, holding your terminal until you press Ctrl-C. Useful for debugging or one-off runs.
start accepts the same two flags. Combine either with PROMPTSTER_DEBUG=1 to print a line per captured event.

Status & diagnostics

Shows your key (masked), the dashboard host, the watched directory, whether the background capture is running and its process ID, your anonymous device ID, and the number of buffered events.
By default it’s a live dashboard that refreshes about once a second. For a single static snapshot — in a script or a non-interactive shell — use any of:
Runs a health check and prints each result with a suggested fix. It confirms:
  • a developer key is present and is a valid PSE- key (and where it came from)
  • your dashboard host is reachable
  • your Claude Code transcript directory exists
Run this first whenever capture seems off. See Troubleshooting.
Prints the installed CLI version.

Internal subcommands

The CLI spawns claude-watch and codex-watch itself to tail Claude Code and Codex transcripts. You don’t run these directly — background capture manages them. They’re mentioned here only so you recognize them if you see them in a process list or in daemon.log.